DE10128147A1 - Transmitting data in computer network involves server using transformation with which representation of resource on client is not altered but data redundancy increased - Google Patents

Abstract

The method involves a client making a request to a server identifying a resource to be transmitted to the client and the server extracting the resource from a data source. The server uses a transformation on the data of the resource with which the representation of the resource on the client is not altered but data redundancy is increased and/or a loss-affected reduction of data quantity achieved. The transformed data are passed to the client. The method involves a client (1) making a request (2) to a server (3), whereby a resource is identified in the request that is to be transmitted to the client for display and the server extracts the corresponding resource from a data source. The server uses a transformation on the data of the resource, at least for defined types of resource, with which the representation of the resource on the client is not altered but the data redundancy is increased and/or a loss-affected reduction of data quantity is achieved. The transformed data are then passed to the client. Independent claims are also included for the following: a server for providing a data transmission service in a computer network, a program for running on a server and a system for accelerating data transmission.

Description

The invention relates to a method for transmitting Data in a computer network and a server as well as a Program for this and a system for accelerating the Da transmission in computer networks.

For the transmission of data in computer networks today for example, the HTTP protocol used by its widespread use in the "World Wide Web" and Has found fame.

This is a protocol for the exchange of Resources (for example HTML files, graphics, sound files en) in computer networks, especially on the Internet. The Exchange begins with a request from the client or browser to a server. This processes the request and sends a response back to the client.

Every request, be it a request or a response, exists of two parts, a head area (Head) and a data area (body). The head area contains additional information about the request, the body area the actual data.

As a typical example, retrieving one as HTML document called "Website" from a www Serve server (world-wide-web). The user who has a sol che page, uses a client program that is referred to as a browser. Various are known Browser, including the "Internet Explorer" from Micro soft or the "Netscape Navigator".

In such a client, the user gives the address of the  Internet resource to be called up as a so-called URL (Uniform Re source locator) a, e.g. B. "http: // server.irgendwo.de/doku ment.html ". The browser now sends as described above a request to a server that is located at "ser ver.irgendwo.de "on the Internet. This request contains the name of the requested document ("doku ment.html "). The request also contains one in the head area Number of information about the request, including informa what character sets are understood, which language is understood etc.

The server accepts the request and answers it with a response. This contains the in her body area desired HTML page ("document.html").

Such an HTML page is a loading description of what the browser should display. The format this description is standardized. Known current ones Standards are HTML and XML. In short, these are based Page description languages that apart from that information is given in what manner this must be displayed (e.g. with regard to size, character set, Bold, color, formatting etc.). This formatting information is given in the form of so-called tags. On An example of such a day is "<b <". This day causes the following text to be bold is printed. The requested HTML document is from the browser interpreted accordingly and displayed on the screen poses.

There are several programs in use today that are called "Web server ", i.e. working as an HTTP service provider. Examples are the Apache web server, the iPlanet web server or the Internet information server from Microsoft. The source Texts from the Apache web server are freely available, so they  can be easily added and expanded. About this The book by Lincoln Stein and Doug MacEachern "Wri Apache Modules in Perl and C "(O'Reilly & Associates Inc., March 1999), ISBN 1-56592-567-X.

When transferring the requested resources is a fundamental problem the transmission speed. This is limited by the band available wide in the transmission of such information, that is The speed at which a quantity of data is which transmission paths can be transmitted. A special one Experience has shown that the bottleneck in Internet connections is Dial-up connection (ISDN or modem) of the user at the Pro vider.

To accelerate the data exchange, the US-A-6021426 proposed several measures. This is what it is about is about caching, data compression and Differential coding.

When compressing data, the size of a file is reduced adorned by using existing redundancies in the data become. A large number of compressions are known to the person skilled in the art known processes, including the Huffman coding, the "De flate "algorithm and the Lempel-Ziv coding Appropriate compression methods are also in the HTTP / 1.1 protocol considered. The header of an HTTP request contains a Field labeled "Accept-Encoding". This field can in addition to the value "identity" (for plain text data) also the Accept the values "gzip", "compress", "deflate" if the Client using an appropriate decompression algorithm disposes.

In a paper by Rendall L. Schwartz, a pearl Script presented a proxy server for HTTP pages  represents. This program evaluates the content of the "Accept- Encoding "field of a request. If from this field it can be seen that the corresponding client compressed Can process data, the program compresses the data and sends the compressed version.

A proxy server is also used in US-A-5918013 wrote. This is an HTTP server that logically between the client (browser) and the respective Servers is arranged. Every request is made by the client to the proxy server, which in turn sends the requested data from the respective server and the forwards the respective response to the client. The response is cached with a "caching proxy" so that in the case of another, identical request, the intermediate memory saved document can be delivered as a response, without having to retrieve the data from the server again.

The proxy server of US-A-5918013 provides that the white redirected document if it contains errors are included. The recoding includes the revision of the Document also for sub troubleshooting pressing unwanted content etc. It will be additional suggested graphics in, for example, from the GIF format transcode the JPEG format because this is a more effective one Compression method is.

In US-A-5931904 a proxy server for HTTP services described the delay time between the request and reduce the response. The procedure described provides two proxy servers, each with different ver hold sions of a document. In cases where this is good is first an older version of the document sent while a newer version is being retrieved. There according to only the differences between the two  Versions submitted. This corresponds to the so-called "Dif reference coding ".

Another proxy server is described in US-A-6049821 ben. This proxy server changes the request as well also the response. The changes include removal of comments, graphics, backgrounds and other than less important elements of the document. The Proxy server takes into account the capabilities of the respective browser based on a list.

It is an object of the invention, the known transmission procedure to improve servers and programs so that the Bandwidth of the connection is used as well as possible.

This problem is solved by a method with the Merk paint of claim 1 (transformation of the data to increase the redundancy and / or transformation of the data for loss affected compression) and a server according to claim 16, a program according to claim 32 and a system according to An Proverb 35. Dependent claims relate to advantageous te configurations.

According to the invention, the server transmits the requested Resource, such as an HTML file, is not necessarily in the same form as it is taken from the data source. But he spends on the resource - preferably only on Resources of a certain type, for example XML or HTML files - a transformation on and then transmitted the result of this transformation (directly or as below described, after using a compression process).

The goal of this transformation is to better compress Availability of the data of the resource or the data reduce the quantity already by applying the transformation  ren. The latter happens "lossy", d. H. that the original version from the transformed version not is more reconstructable.

In any case, however, the transformation is applied ensures that this will render the resource is not changed on the client. That means that by the transformation is the information content of the data is changed, but this change for the user who not the source text data, but only the representation considered to the client, is not recognizable.

Such a change is particularly important with HTML or XML Files possible. With these page description languages parts of the source text can be changed without the - corresponding to the standard - representation of the corresponding the sides is changed.

The word "program" is used here in the broadest sense. This includes both codes in directly executable dimensions language, for example the product of a compiler Run. The term "program" also includes Script languages such as B. perl or other interpreter language Chen or for example Java applets.

"Server" means a computer that is in a Network offers a network service. Preferably poses the server has an HTTP service available.

A "client" is understood to mean a computer that is in is able to connect to the server over the network in one of the Client / server architecture known to those skilled in the art process, the client requesting corresponding requests to the server ver and the server responded with a response words. Well-known browser programs are examples of clients  me for displaying HTML pages.

An HTTP service is understood to mean a service that is based on an appropriately formatted request Response sent to a client. This includes the current standardization of the HTTP protocol and its successor or modifications.

In the data source used according to the invention, from which the requested resource is taken from, at for example, a simple data storage, e.g. Legs Hard disk (non-volatile memory) or, for frequent pages called up, the main memory (RAM memory, curs act memory) of the server computer itself. The HTTP server can also work as a proxy server, i.e. H., that he in turn sends a request to a request another server and this as a data source uses. For example, there is also a data source installed on the server computer or with it on the network connected database in question, from the z. B. dynamic genes content, such as HTML pages in response to a search query.

The data of the response are preferred in the data source Taken in plain text, which unlike some today Known server implementations, each of which HTML file a normal and a compressed version the hard disk and the server depending on Client's skills either one or the other Version shipped.

The Trans. Applied to the resource according to the invention formation has two aspects, alternatively or in Kombina tion are used. On the one hand, the Trans formation, the redundancy of the plain text data can be increased.  

This will compress the resource's data better bar. This is based on the observation that all compressors proceed the redundancy existing in a data set use to compress the data. The degree of compression, d. H. the ratio of the original data size to the size the compressed version of the data is all the more special The greater the redundancy in the data.

An example of redundancies in files is when the Data contains a number of repetitions. Instead of that save repeated sequences completely, a suitable compression method can le each sequence Save only once and then only the number and Mark the position of the respective repetitions. Correspond appropriate compression algorithms (e.g. gzip, deflate, etc.) are known to the person skilled in the art.

Another example of redundancy is if not all characters available in a character set with occur at the same frequency. In this case, a Frequency analysis can be done, often before coming characters by short, rarely occurring characters are represented by long symbols. Also here are corresponding procedures (e.g. Huffman coding) to the subject man known.

The transformation routine that provides the redundancy of the plain text data should increase, z. B .. at these points start and carry out a transformation in which the Number of repetitions in the file are increased or the frequency distribution of the individual occurring book is changed.

Such a trans is particularly useful for HTML or XML files  formation advantageous, the HTML or XML source text is changed, but the display on the client is the same remains. For example, some or all of you can use ten tags are written as lowercase letters. Hereby On the one hand, the redundancy of the data is increased by a higher number of repetitions (some tags occur very often often on), on the other hand by a less uniform Frequency distribution (lowercase letters occur considerably more often than capital letters).

The transformation can also be called whitespace elements to edit. These are elements of the sign sentence that act as spaces and often forma source code can be used. Count among them the ASCII symbols space (), tab (), form feed (), zero-width space () and lines breaks. These different symbols are in source code Often inserted for readability. At the Darstel HTML or XML pages play the different Whitespace elements don't matter, however, it does (if at all) just insinuate that a Whitespace element exists. To increase redundancy can therefore always use the different whitespace elements by one and the same element, for example ASCII space be replaced. This affects the frequency distribution and leads to a larger number of repetitions, so that redundancy increases.

Another possible transformation with which redundancy can be increased without the display of an HTML or XML page is thereby changed, is the replacement equivalent tags. In HTML, for example, are the tags "<b <" (for "bold", bold) and "<strong <" equivalent. If all "<strong <" tags are replaced with "<b <" tags, on the one hand there is a reduction in the amount of data because  the "<b <" day is shorter. In this way, the ver different tags always by one and the same (equivalent gen) day replaced. This increases the redundancy if the Day occurs frequently and therefore a corresponding number Repeats.

In terms of programming, such a replacement can be advantageous via a lookup table, d. H. a replacement table reali be settled. After a day of going through the plaintext is recognized, this is looked up in the table. The The table then contains the equivalent replacement tag that is on the appropriate position should be set. So can a replacement of umlauts, e.g. B. "ä" through the HTML code Correspondence () can be made.

Some HTML tags can contain arguments (for example "<a href = adresse.irgendwo.de <"). There are no other tags Arguments (for example, "<title <"). Made of performance green it can make sense to only include HTML tags without arguments to handle the transformation. These can be recognized safely and the implementation of these tags is in lower case also harmless with certainty. With arguments, however, is in some cases the transformation of the arguments is not wishes (e.g. for file names where, depending on the company system between upper and lower case will be divorced).

The second aspect of the transformation according to the invention, the alternatively or cumulatively to increase redundancy can be used affects a lossy compression tion of the plain text data by which the data is presented is also not changed.

After the lossy compression, the length of the transformed plaintext less than the length of the original  original language. Here are parts of the plain text removed so that - unlike compression routines such as B. gzip - actually a change in plain text data takes place in which from the transformed data the original data cannot be reconstructed (because lossy). However, only one is ver lusty compression applied to the representation the resource, for example an HTML page, is not affected.

An example of such a transformation is replacement sequences of successive whitespace elements, e.g. B. several consecutive ASCII space or Kombina functions of whitespace elements (e.g. a line break, then several tab elements) through a single ASCII space. This implementation reduced the length of the HTML page but no impact on the presentation, provided the Inter pretation by the browser is standard-compliant.

HTML and XML source codes can also be created using special tags tagged comments. These comments fin not taken into account in the presentation. You can therefore without replacement for lossy compression be deleted.

Increasing the redundancy of the plain text data enables one better data compression and thus better utilization the bandwidth. This applies even without the direct application of Data compression algorithms by the server itself. Because a number of transmission paths, including modem lines, already use compression methods, which increase from the Redundancy benefits. But is particularly advantageous a development of the invention, according to which the request (be preferably the header information of the request) is evaluated, to determine whether the client is responding in a co processed form. This can be  for example about the "Accept-Encoding" field of an HTTP request act. If this field has the value "gzip", "compress" or "deflate", the client can use the appropriate algo Process compressed data. If this is recognized the server program uses the appropriate coding step, for example an appropriate encryption, but preferably a known compression method the transformed data. Here it has a special effect otherwise advantageous if previously through the transformation the redundancy of the plain text data has been increased. Because everyone Known compression methods take advantage of the redundancy in order Compress data. With increased redundancy, the com degree of compression higher, so that the available tape width of the transmission path is better exploited and the Resource transfer faster, d. H. in shorter time, is completed.

The implementation of a method according to the invention or of a server according to the invention is usually performed by a computer acting as a server, the one executes the corresponding program. At the server computer it is for example a conventional PC (Personal Computer) with common hardware components, d. H. a central unit with one or more processors, RAM, a hard drive (or a comparable one Mass storage) act.

As a possible implementation of parts of the Server Pro grammes, for example the coding algorithm, can Implementation in an interpreter language. However, it is preferred that the program is completely compi can be d. H. after a translation process as direct executable code is complete.

It is also preferred that the program complete the coding step  (e.g. application of the compression process) "on the fly" performs, d. H. immediately when calling a clear text res source of which created a code version in the working memory and then this code reading from the main memory averages. Dispensing with file access speeds up the process the execution of the corresponding program.

A conventional one can already exist on the server computer Server program (e.g. the "Apache" mentioned above) as the original jump server expire. The server program that according to the Invention works and used as an acceleration server can then also run on the same computer, operating as a proxy server, i.e. H. the conventional, server program running in parallel as a data source.

However, the server program according to the invention is preferred on a dedicated computer (dedicated Server) to run. When used as a proxy Server can thus an existing system with an origin Servers can be redesigned with little effort so that an acceleration server is also available. A sol cher server computer can be a complete product, best based on hardware and software and existing systems with minimal configuration effort be "upstream" for acceleration.

Remarkable when using an accelerator according to the invention acceleration server is the position that the acceleration Server in the entire system. As mentioned serves the registered transformation of the better utilization of the Bandwidth, especially through additional compression ren. This acceleration affects the "downstream" area, d. H. the transfer from the (acceleration) server to the client.

Proxy servers are always logical between the client and (Ur  jump) server arranged. While conventional proxy However, servers tend to be in the range of one or more Clients (e.g. with an ISP, i.e. an Internet access provider) are arranged and via this proxy server Access to all possible servers is routed to Acceleration server preferably a single one narrowly assigned to a group of servers. The acceleration The cleaning server then serves as a connection for all possible clients running place and accelerated by the applied transforma tion and possibly compression the downstream traffic for the resources retrieved from the origin server.

This arrangement is a logical arrangement, i. H. the physi Kalische standards can deviate from this. But in the In practice, it is also recommended to use the acceleration server to be placed close to the origin server, because this is preferred both via a broadband connection (i.e. broadband as a dial-up connection, e.g. B. via modem or ISDN before from 100 Mbit / s) are coupled. Broadband network Connections are made over short distances, e.g. B. within one Building, very inexpensive to implement (e.g. via Twi sted-pair power cable, 100 or 1000 Mbit / s, or via direct Fiber optic connections).

An embodiment of the invention will be hereinafter wrote. For this purpose, reference is made to the drawings. The drawings show:

Figure 1 is a diagram of the flow of data exchange according to the HTTP protocol.

Fig. 2 is a diagram of structure of a HTTP request;

Fig. 3 is a diagram showing the configuration of a HTTP response;

. Figure 4 is a symbolic representation of the components of a system with an HTTP server;

Fig. 5 is a flowchart of the response to a request by the server;

Fig. 6 is a symbolic representation of the components of a system of origin and acceleration server.

Fig. 1 illustrates the sequence of exchanging data according to the HTTP protocol. An HTTP server 3 (Web server) provides the corresponding service. The user uses a client 1 (browser) to retrieve data from the server 3 . The client 1 formulates a request 2 , which is transmitted to the server 3 . The server 3 formulates a corresponding response 4 , which is returned to the client. The client takes the desired information from Response 4 . In the case of an HTML document, the client interprets the corresponding instructions for displaying a page and makes them visible to the user.

Fig. 2 shows the structure of a corresponding Requests 2. This consists of a head and a body area. The header area contains various fields, including the "Accept-Encoding" field and other fields, for example "Ac cept-Language". The information contained in these fields tells the server in what form the client can use the requested information. In the example shown in FIG. 2, the content of the "Accept-Encoding" field is the token "gzip". A set of corresponding tokens is defined in the HTTP1.1 specification. Here "gzip" means a compression format that is produced by the compression program "gzip" (GNUzip), which is described in RFC1952. It is a Lempel-Ziv coding method (LZ77) with 32 bit CRC (Cyclic Redundancy Check).

FIG. 3 shows an exemplary response 4 to the request from FIG. 2. Response 4 has a head and a body area. The header area contains various fields, including the "Content Encoding" field. In the example shown in this field, this field contains the token "gzip", which means that the data contained in the body was compressed using a "gzip" method as desired in the head area of the request. Correspondingly, the body area of the response then also contains the data compressed with this method.

In FIG. 4, the hardware-Kom involved in the above-described data transmission are in a symbolic representation illustrated components. On the one hand, these include a server computer 10 and a number of client computers 12 . These computers are connected to one another via a data network 14 , generally the global Internet. The clients 12 are distributed in a wide variety of locations. They are coupled to the data network 14 via various transmission paths 16 , including modem lines, dedicated lines etc. The coupling of the clients to the network is generally narrowband with bandwidths of approximately 33 kbit / s to 64 kbit / s.

The client computers 12 can be very different devices, including workstations, personal computers, internet-capable telephones, etc. A client program is executed on the client computers 12 , for example Microsoft Internet Explorer, Netscape Navigator or others. These programs work as an HTTP client as described above. You can display the resource retrieved from the server computer 10 via the HTTP protocol for the user. If the resource is an HTML or XML file, it is interpreted by the browser program and displayed accordingly.

Server computer 10 can also be any type of computer. However, servers that process a large number of requests usually have significantly better equipment than the average client computer. Systems with several processors and high processing speed as well as a large working memory are common here.

The server computer 10 is connected to the Internet via a network connection 18 . The network connection 18 is relatively broadband with typical data transmission rates of at least 1 Mbit / s.

The server computer 10 runs an HTTP server program. Such programs are known to the person skilled in the art, for example the Apache HTTP server, the iPlanet web server or the Internet Information Server from Microsoft. The program running on the server computer 10 of the exemplary embodiment behaves essentially like conventional HTTP servers, ie it responds to requests from clients by removing the requested resources from a hard disk 11 and transmitting them as a response to the client. The decisive difference, however, is the application of a special transformation routine to the data of certain retrieved resources and the subsequent compression of the data transformed in this way, if the respective client can process this compression.

In the following, an implementation of the program running on the server computer 10 is presented and explained using the flow chart of FIG. 5. The program written in the programming language C is attached; in the explanation, reference is made to the line numbering there. The corresponding routine is intended for integration into the iPlanet server from Netscape.

When the web server 3, shown as a large box in FIG. 5, receives an HTTP request 2 from the browser 1 , it first checks the type of the requested resource. If it is an HTML or XML file, the "accele rate" function (from line 74) is called. Otherwise the resource will be transferred as usual.

The "accelerate" function initially performs a number of Checks by whether the request is valid and error-free is (line 133-170).

Then the data is prepared (line 178 ff). Here two transformations are applied to the data. Egg on the other hand, all "simple" tags (i.e. without arguments) converted to lower case. On the other hand, will be on successive whitespace elements by a single he puts.

The corresponding routine contains a "tag_start" flag. Is If this is set to the value 0, the current Pointer position not processed a day. Otherwise tag contains start the position of the start of the currently edited tag.

The entire plaintext file buffer is looped down Run through line 190 with an index i and one transformed version input where the index fsize points to the end. If here sequences successive whitespace elements are encountered, is only in the transformed version (input [fsize]) only single space saved (line 194-201). Otherwise the content of the plain text data (buffer) is initially identified table in the transformed version (line 205).

If the character "<" is found, the flag tag start set to the current index position fsize. If a "<" is found in the entire transformed version since the tag_start flag was set, the text content of the tag  converted to lower case (line 217).

Appears while running through the input file buffer () Character found that is neither alphanumeric nor a "/" (Line 221), the flag tag_start is set to zero. As a result, tags with arguments are not converted because there is always a character in these tags that this condition is not met (e.g. a whitespace).

Then the field "Accept-Encoding" of the header ches of the request evaluated (line 232). If no com is possible, the data will be uncompressed sends (line 240), otherwise a CRC 32 checksum is loaded calculates (line 277) and the transformed data input are compressed using the deflate algorithm (line 284- 286). The well-known zlib program library used (line 23), the appropriate compression routines.

The response is then sent to the Client prepared by first using the appropriate gzip header is written (line 300-307). The so prepared The response is returned to the browser.

Using this special HTTP server program, the data transmission can initially be carried out according to the HTTP protocol as is conventionally known, ie a client 12 transmits a request 2 to the server computer 10 via the transmission paths 16 and 18 and the data network 14 , where a specific desired resource is specified in the request. The server computer 10 accesses its hard disk 11 in order to obtain the desired resource and transmits it back to the client computer 12 in a response 4 .

Only if the requested resource is an HTML or XML document does the client 12 - depending on his capabilities in compressed or uncompressed form - not necessarily receive the resource in the same form as it is on the hard disk 11 of the server. Computer 10 is stored. Instead, due to the lossy compression and increase in the redundancy within the document, the client 12 receives a transformed version, the representation of which, however, does not differ from the original version of this document in accordance with the specification. The transformed version is already shorter than the original one or can be compressed better, so that overall the bandwidth of the transmission path over the network connections 16 , 18 and the data network 14 can be better utilized. This applies in particular to the network connection 16 of the client computer 12 with an Internet service provider, if this is a modem connection. This is because all current modem protocols for data transmission include a (lossless) compression process, in which the degree of compression becomes cheaper with higher redundancy.

In FIG. 6, a second embodiment of the invention is shown. In this embodiment, too, a number of client computers 12 , each running a browser program, are connected to the data network 14 via (narrowband) network connections 16 . Here, too, there is a data connection 18 from the symbolically represented provider location 20 (for example a building, data center, etc.) to the data network 14 . Within this provider location 20 , an origin server 22 and an acceleration server 24 are connected to the network connection 18 . The origin server 22 executes a conventional HTTP server program. If requests are made to this HTTP server, it retrieves the resources specified therein from the hard disk memory 23 .

In addition, there is an acceleration server 24 within the operator location 20 , which is also connected to the network connection 18 . The acceleration server 24 is a stand-alone computer that is the only program that executes an HTTP proxy server program, which will be described in more detail below.

The acceleration server is, for example, a personal Computer with four high-frequency processors. He is with fast RAM in sufficient quantity, for example 256 MB, equipped. That comes as the operating system very good network-compatible Linux.

The program running on the acceleration server 24 is basically an HTTP proxy server. In contrast to conventionally known proxy servers, to which no usual HTTP request is sent, but which require special proxy requests and which, based on this request and the origin server specified therein, send a request to these, the HTTP behaves Accelerator program to the outside like a real HTTP server. If a certain resource of this HTTP server is queried, the HTTP acceleration program does not take it directly from a local data source, but instead sends an HTTP request to the origin server 22 , which for example retrieves the resource from its hard disk 23 removed and transmitted to the acceleration server 24 as a response. This forms the response to the original request.

However, the proxy computer does not simply forward the resource obtained in this way. Instead, he applies the transformation presented in connection with the first exemplary embodiment to HTML and XML resources. Here, a program routine is used which is very similar to the routine described above and shown in FIG. 5. The only difference is that the resource is not obtained directly from the data source 23 , but that it is queried by the origin server 22 .

The system of origin server 22 and acceleration server 24 is configured such that an HTTP request directed to the provider's network address is first processed by the acceleration server. The acceleration server obtains the requested resource from the origin server 22 , applies the transformation described above to HTML and XML files and compresses them if the client has specified in the request that compressed contents can be processed. This division of tasks between acceleration server 24 and origin server 22 is not visible to the client.

With the help of a separate acceleration server 24 , existing systems can be accelerated very easily without deep intervention being necessary. In addition, no additional computing load is generated on the original server 22 by the application of the transformation and implementation of the compression method.

Although the system is shown in FIG. 6 in such a way that the origin server 22 can also be accessed directly from the data network 14 , this is not absolutely necessary. Even if the physical connection corresponds to the representation in FIG. 6, the servers 22 , 24 and programs running thereon can be configured such that the acceleration server 24 is logically always arranged between a client 12 and the originating server 22 . Of course, such an arrangement can also be implemented physically by equipping the acceleration server 24 with two network interfaces.

In principle, the application of the acceleration server 24 to the origin server 22 will be so broadband that there is no delay in the data transmission here. If these - as indicated in Fig. 6 - see together at a common location, for example in the same building - or even directly next to each other - a broadband connection, for example via an optical network, can also be implemented simply and inexpensively.

In the modification, the use of an acceleration server is also possible as a pure software solution, ie as a separate program on the origin server computer 22 .

Claims (37)

1. Method for transmitting data in a computer network ( 14 ), in which
a client ( 12 ) makes a request ( 2 ) to a server ( 10 , 24 ),
wherein the request designates a resource that is to be transmitted to the client ( 12 ) for display
and the server ( 10 , 24 ) takes the corresponding resource from a data source ( 22 , 11 ),
characterized in that
the server ( 10 , 24 ) at least for predetermined types of resources
applies a transformation to the data of the resource, but does not change the representation of the resource on the client ( 12 )
the redundancy of the data is increased
and / or a lossy reduction in the amount of data is achieved,
and then the transformed data is transmitted to the client ( 12 ). 2. The method according to claim 1, wherein
the request ( 2 ) from the client ( 12 ) is then evaluated as to whether the client ( 12 ) can process the resource in coded form,
and in this case a corresponding coding is carried out after the transformation,
and the encoded data are transmitted to the client ( 12 ). 3. The method according to claim 2, wherein the coding Data compression method includes. 4. The method according to any one of the preceding claims, wherein the data exchange between the client ( 12 ) and server ( 10 , 24 ) takes place according to the HTTP protocol. 5. The method according to any one of the preceding claims, at with the predetermined types of resource files Page description languages, especially HTML files or XML files. 6. The method according to any one of the preceding claims, wherein the data source ( 11 ) is a local non-volatile or volatile memory of the server computer ( 10 ). 7. The method according to any one of claims 1 to 5, in which
the data source is a database,
whereby the resource is generated dynamically. 8. The method according to any one of claims 1 to 5, in which
the server ( 24 ) works as a proxy server
the data source being an origin server ( 22 ),
and the proxy server ( 24 ) makes a request to the original server ( 22 ) so that it transmits the desired resource. 9. The method according to claim 8, wherein
the proxy server ( 24 ) processes HTTP requests directly,
and obtains the resources for all requests from only one origin server ( 22 ). 10. The method according to any one of the preceding claims, in which
the transformation includes the treatment of whitespace elements,
whereby individual whitespace elements and / or groups of successive whitespace elements are always replaced by one and the same whitespace element. 11. The method according to any one of the preceding claims, in which
the transformation includes the conversion of tag elements into uniform upper / lower case, especially lower case of all letters. 12. The method of claim 11, wherein the transformation only applied to tag elements without arguments. 13. The method according to any one of the preceding claims, in which
certain tag elements are replaced by equivalent replacement tag elements,
in particular where the replacement tag elements are shorter than the tag elements to be replaced. 14. The method of claim 13, wherein the replacement over a replacement table is controlled. 15. The method according to one or more of the preceding Claims where the transformation is the distance covered by comments. 16. Server for providing a data transmission service in a computer network ( 14 )
at least one network interface for coupling to a computer network ( 14 )
Via which the server ( 10 , 24 ) accepts requests from a client ( 12 ) from the computer network ( 14 ), in which a resource is designated that is to be transmitted to the client ( 12 ) for display
and at least one connected data source ( 11 , 22 ) from which the server ( 10 , 24 ) takes the corresponding resource,
marked by
Means for carrying out a transformation which the server ( 10 , 24 ) applies to at least predetermined types of resources, in which the representation of the resource is not changed by the client ( 12 ),
whereby the redundancy of the data of the resource is increased by the transformation and / or a lossy reduction in the amount of data is achieved,
and means for transmitting the transformed data to the client ( 12 ). 17. The server of claim 16, wherein
the server ( 24 ) is a computer with a central unit and a working memory,
the central unit executing a server program that works as a proxy server,
an origin server ( 22 ) serving as the data source,
from which the server ( 24 ) requests the resource. 18. The server of claim 16 or 17, wherein
the request from the client ( 12 ) is evaluated to determine whether the client ( 12 ) can process the resource in coded form,
and in this case a corresponding coding is carried out after the transformation,
and the encoded data are transmitted to the client ( 12 ). 19. The server of claim 18, wherein the encoding Data compression method includes. 20. Server according to one of claims 16 to 19, in which the data exchange between client ( 12 ) and server ( 10 , 24 ) takes place according to the HTTP protocol. 21. Server according to one of claims 16 to 20, wherein the predetermined types of resource files with pages description languages, especially HTML files and XML files are. 22. Server according to one of claims 16 to 21, wherein the data source is a local volatile or non-volatile memory ( 11 ) of the server computer ( 10 ). 23. Server according to one of claims 16 to 21, in which
the data source is a database,
whereby the resource is generated dynamically. 24. Server according to one of claims 16 to 21, in which
the server ( 24 ) works as a proxy server,
the data source being an origin server ( 22 ),
and the proxy server ( 24 ) makes a request to the original server ( 22 ) so that it transmits the desired resource. 25. The server of claim 24, wherein
the proxy server ( 24 ) processes direct HTTP requests,
and obtains the resources for all requests from only one origin server ( 22 ). 26. Server according to one of claims 16 to 25, in which
the transformation includes the treatment of whitespace elements,
whereby individual whitespace elements and / or groups of successive whitespace elements are always replaced by one and the same whitespace element. 27. Server according to one of claims 16 to 26, in which
the transformation includes the conversion of tag elements into uniform upper / lower case, especially in lower case of all letters. 28. Server according to claim 27, in which
the transformation is only applied to tag elements without arguments. 29. Server according to one of claims 16 to 28, in which
certain tag elements are replaced by equivalent replacement tag elements,
especially where the replacement tag elements are shorter than the replaced tag elements. 30. The server of claim 29, wherein the replacement is via a replacement table is controlled. 31. Server according to one of claims 16 to 30, wherein the Transformation involves the removal of comments. 32. Program to run on a server according to one of the  Claims 16 to 31, wherein the program is a method according to one of claims 1 to 15. 33. The program of claim 32, wherein
the program is fully compilable. 34. Program according to one of claims 32 or 33, wherein
the program works entirely in volatile memory. 35. System for accelerating data transmission in computer networks,
an origin server ( 22 ) and a number of clients ( 12 ) are present with the computer network ( 14 ),
and the clients ( 12 ) make requests,
a resource is designated in the inquiries, which is to be transmitted to the client ( 12 ) for display,
and the origin server ( 22 ) provides the appropriate resource
characterized in that
an acceleration server ( 24 ) is additionally provided, which is logically connected upstream of the origin server ( 22 ),
the requests are processed by the acceleration server ( 24 ),
by first sending a procurement request to the origin server ( 22 ) so that it transmits the desired resource,
and by the acceleration server ( 24 ) at least for predetermined types of resources
applies a transformation to the data of the resource which does not change the representation of the resource on the client ( 12 ), however
the redundancy of the data is increased and / or a lossy reduction in the amount of data is achieved,
and then the transformed data is transmitted to the client ( 12 ). 36. The system of claim 35, wherein
the acceleration server ( 24 ) is only connected directly upstream of an origin server ( 22 ). 37. The system of claim 35 or 36, wherein
a high bandwidth connection connects the acceleration server ( 24 ) to the origin server ( 22 ).