20. September 2012

The Day We Invented AJAX and REST

While searching for prior art for Eolas' claims, that the U.S. Patent No. 7,599,985 covers AJAX, I just discovered this:

AJAX in June 1998

The CoBrow group at Distributed Systems Dept., Ulm University did AJAX style communication in June 1998.

In '98 we implemented a HTML only client for the CoBrow project. This means a dynamically changing HTML page, which shows updated data whenever things changed on the server. CoBrow was about meeting people on web pages. A HTML based chat was used to show a Java applet free dynamic GUI.

Today you'd say: dynamic HTML and server updates? no problem. Use jQuery, AJAX, WebSocket, maybe even XMPP and BOSH for the routing. But in 98, there was no AJAX, no XMLHttpRequest, not even iframe for all browsers (IE and Netscape 4).

So, the software engineer uses what's available in all browsers:
  • a frameset with frames and 
  • a HTML form which can be 
  • loaded from a server and 
  • read by other frames.
The request-frame keeps loading a HTML page with a form from the server. The form contains named form fields. The GUI-frame read the fields and uses the data to show something. There are no cross-domain issues, because the GUI-frame and request-frame have the same domain. A simple way to request key-value pairs from the server.

Source Code

Now comes the tricky part. At that time we programmed web pages with shell/perl scripts. There was not very much else, except for PHP 2, which is not the PHP you want. So much for stateless pages. No Java server runtimes, no application servers.

If you wanted persistence, multiple concurrent clients and other fancy stuff in the server, then you implemented your own server including the HTTP parser. Therefore, the server was written in plain C with a kind of object oriented style. Just as other guys did at the same time, e.g. the apache devs Brian Behlendorf, Roy T. Fielding et al. Could have been C++, though, but wasn't.

The feature was "checked in" to the code base on June 22nd, 1998.

On Mar 29, 1999, the Cobrow Newsletter - No. 3 announced the "HTML user interface":
http://www.virtual-presence.org/external/www.cobrow.com/pages/docs/cobrow-news-3.txt

For those of you interested in software archaeology, here comes the code from CoBrow Server "cbVici", version 2.1.100:

Development of the "HTML user interface" started:

Print.c(310):
  PrintHistoryItem(conn,  style, "22.05.1998", "1.13.6", "hw", "Split off Primitives from Command, added module HtmlUI");

AJAX like function added 1 month later:

Print.c(308):
  PrintHistoryItem(conn,  style, "22.06.1998", "1.14",   "hw", "response=application/x-html-formencoded response data");

Implementation:

Response.c(363):
#define RESPONSECODER_FORM_HEAD "<HTML>\n<BODY BGCOLOR=\"#%s\">\n<FORM name=\"%s\">\n"
#define RESPONSECODER_FORM_ITEM "<INPUT name=\"%s\" type=\"hidden\" value=\"%s\">\n"
//#define RESPONSECODER_FORM_ITEM "<INPUT name=\"%s\" type=\"text\" value=\"%s\"><br>\n"
#define RESPONSECODER_FORM_TAIL "</FORM>\n</BODY>\n</HTML>\n"

Response.c(524):
int CBResponseCoder_Add(ResponseCoder *coder, char *name, char *value)
{
  int ok= false;
  if (coder == NULL) return ok;

  if (! strcasecmp(coder->codertype, QUERYARG_RESPONSE_VALUE_FORM)) {
    if (coder->cnt+ strlen(name)+ strlen(value)+ strlen(RESPONSECODER_FORM_ITEM) +1 < coder->size) {
      sprintf(&(coder->buf[coder->cnt]), RESPONSECODER_FORM_ITEM, name, value);
      coder->cnt+= strlen(&(coder->buf[coder->cnt]));
      ok= true;
    } else {
      TRACE(TR_ERR, "CBResponseCoder_Add: RESPONSECODER_MAX_ITEM_LEN exceeded");    
    }
  } else if (! strcasecmp(coder->codertype, QUERYARG_RESPONSE_VALUE_JSCOOKIE)) {
    if (coder->cnt+ 64+ strlen(name)+ strlen(value)+ strlen(RESPONSECODER_JSCOOKIE_ITEM) + strlen(gMyDomainName) + 1 < coder->size) {
      char gCookieDomainName[MAX_HOST_NAME_SIZE];
      sprintf(&(coder->buf[coder->cnt]), RESPONSECODER_JSCOOKIE_ITEM, name, value, assert2dots(gCookieDomainName, gMyDomainName));
      coder->cnt+= strlen(&(coder->buf[coder->cnt]));
      ok= true;
    } else {
      TRACE(TR_ERR, "CBResponseCoder_Add: RESPONSECODER_MAX_ITEM_LEN exceeded");    
    }
  } else if (! strcasecmp(coder->codertype, QUERYARG_RESPONSE_VALUE_KEYVALUE)) {
    if (coder->cnt+ strlen(name)+ strlen(value)+ strlen(RESPONSECODER_KEYVALUE_ITEM) + 1 < coder->size) {
      sprintf(&(coder->buf[coder->cnt]), RESPONSECODER_KEYVALUE_ITEM, name, value);
      coder->cnt+= strlen(&(coder->buf[coder->cnt]));
      ok= true;
    } else {
      TRACE(TR_ERR, "CBResponseCoder_Add: RESPONSECODER_MAX_ITEM_LEN exceeded");    
    }
  } else {
    if (coder->cnt+ strlen(value)+ 1 < coder->size) {
      sprintf(&(coder->buf[coder->cnt]), RESPONSECODER_PLAIN_ITEM, value);
      coder->cnt+= strlen(&(coder->buf[coder->cnt]));
      ok= true;
    } else {
      TRACE(TR_ERR, "CBResponseCoder_Add: coder->size exceeded");    
    }
  }

  return ok;
}

Command.c(1203):
 if (!strncasecmp(subsubsection->part, SUBSUBSECTION_PROPERTIES, strlen(SUBSUBSECTION_PROPERTIES))) {
// GET /users/<id>/properties/<propertyname>
  PropertyList *propertylist= NULL;
  //                  ListHdr cmd_params;
  //                  ResponseCoder *coder= NULL;
  //                  ClearList(&cmd_params);
  //                  StringPartArgs2List(theJob->args, &cmd_params);

  if (subsubsection->next != NULL) {
   name= subsubsection->next->part;
   TRACE(TR_CMD, "GET /%s/%s/%s/%s\n", notnullptr(section->part), notnullptr(subsection->part), notnullptr(subsubsection->part), notnullptr(name));
   if ((propertylist= CBClient_GetProperties(theClient)) != NULL) {
    value= CBProperty_FindPropertyInList(propertylist, name);
    if (value != NULL) {
     Elem *param= NULL;
     if (FindCaseNamedElemInList(CMD_PARAM_RESPONSE, (Elem**) &param, &cmd_params)) {
      coder= CBResponseCoder_New(ReadStrElem(param));
     } else {
      coder= CBResponseCoder_New("");
     }
     if (coder != NULL) {
      CBResponseCoder_Open(coder, "get_single_property_response");
      CBResponseCoder_Add(coder, name, value);
      CBResponseCoder_Close(coder);
      JobResponse(theJob, HTTP_OK, CBResponseCoder_GetType(coder), CBResponseCoder_GetData(coder), CBResponseCoder_GetLength(coder));
      CBResponseCoder_Dispose(coder);
     } else JobResponse(theJob, HTTP_INTERNALERROR, HTTP_CONTENT_HTML, "CBResponseCoder_New failed", 0);
    } else JobResponse(theJob, HTTP_NOTFOUND, HTTP_CONTENT_HTML, "no such property available", 0);
   } else JobResponse(theJob, HTTP_NOTFOUND, HTTP_CONTENT_HTML, "no properties available", 0);
...

See how the client could control the encoding by requesting a specific MIME-type: application/x-html-formencoded. HTML-formencoded response data, just great.

 I do not imply, that we invented AJAX and everyone just copied. At the time, there were other groups with similar ideas, including the IE dev team at MS. But in June 98 we definitely invented client-server requests for data independently of other groups at a very early date.

The source code was publicly available. It was announced by an email newsletter. What else can be done? That's what I call a state of the art and prior art with respect to U.S. Patent No. 7,599,985.

Thanks to contributing coders, department heads and project leads:
  • Heiner Erne
  • Axel Busch
  • Holger Boenisch
  • Lars Kaufmann
  • Dieter Finkenzeller
  • Konrad Froitzheim
  • Peter Schulthess 
_happy_ajaxing()

BTW: the REST part is in Command.c(1203): e.g. URIs like /users/<id>/properties/<propertyname>

Just as if AJAX-like client-server data requests are not enough. They also use REST-like URIs where the URI identifies data resources and not just a SOAP-like service access point.

Those were the days.

Labels: , ,

Keine Kommentare:

Kommentar veröffentlichen

Abonnieren Kommentare zum Post (Atom)