Re: Submitting input-form data to server

ebina@ncsa.uiuc.edu
Message-id: <9309060522.AA20418@void.ncsa.uiuc.edu>
To: "Tony Johnson (415) 926 2278" <TONYJ@scs.slac.stanford.edu>
Cc: marca@ncsa.uiuc.edu, wei@sting.berkeley.edu, hackers@ora.com,
        putz@parc.xerox.com, www-talk@nxoc01.cern.ch, robm@ncsa.uiuc.edu
Subject: Re: Submitting input-form data to server 
In-reply-to: Your message of "Sun, 05 Sep 93 13:31:48 PDT."
             <01H2LAD219KYBHA97R@SCS.SLAC.STANFORD.EDU> 
Date: Mon, 06 Sep 93 00:22:04 -0500
From: ebina@ncsa.uiuc.edu
Status: RO

>  
>  In the sub-group that discussed this issue at W5 we had concluded that the 
>  method of sending the form response as part of the URL would not really 
>  scale elegantly in the future. Consider the form:
>  
>  Your name:<INPUT name="address"><p>
>  5000 word essay describing why you should win the free trip to Australia:
>  <input name="essay" size = 80x500>
>  
>  This would clearly result in monstrous URL's, yet it is not 
>  unrepresentative of some potential uses of forms.

Burn me for a heretic, but there is no significant difference scaling-wise
between GET <url>?essay=<5000 words> and SUBMITFORM <package including 5000
word essay>.  Both send a command, and then 5000 words over the net.

It is not a monstrous URL, because the URL is only everything up to the
'?', everything else is the form data.

If you want a separate command like SUBMITFORM because it is more logical
than overloading GET, thats fine, but lets use real reasons.
>  
>  Since HTTP/1.0 would support new methods in addition to GET, a 
>  reasonable possibility seemed the addition of a new method (say SUBMITFORM)
>  for submitting completed forms. Maybe the reply method should be something
>  that can be specified ACTION modifier of the <FORM> tag, with the possibilities 
>  being 
>  
>  mailto:<e-mail address>     Form sent as e-mail to specified recipient
>  http:<address>              Indicates to return form data as a query string
>  submitform:<address>        Indicates to submit the form using the SUBMITFORM
>                              method.
>  
>  Comments?

All that said, I fully agree that ACTION will need to be extended to allow
multiple methods of form submission.

>  
>  Tony
>  
>  
>  

	Eric