Re: comments on HTTP draft of 5 Nov 93

Tony Sanders <sanders@bsdi.com>
Errors-To: sanders@bsdi.com
Errors-To: sanders@bsdi.com
Message-id: <199312082014.OAA01036@austin.BSDI.COM>
To: Jim Davis <davis@dri.cornell.edu>
Cc: www-talk@nxoc01.cern.ch
Subject: Re: comments on HTTP draft of 5 Nov 93 
In-Reply-To: Jim Davis's message of Wed, 08 Dec 1993 13:30:42 EST.
Errors-To: sanders@bsdi.com
Reply-To: www-talk@nxoc01.cern.ch
Organization: Berkeley Software Design, Inc.
Date: Wed, 08 Dec 1993 14:14:36 -0600
From: Tony Sanders <sanders@bsdi.com>

> But it would seem that SPACEJUMP and TEXTSEARCH are not "methods"
> at all, but rather a pair of keywords that inform us of the
> capabilities of a particular server or document.  
> 
> The meaning of these keywords is that server is prepared to
> accept URLs with particular syntaxes and interpret them according
> to a set of shared conventions.  
Agreed.  We need an extensible attribute scheme.  Of course, you could
just use headers but that tends to be too unstructured.

> It seems to me that TEXTSEARCH means ISINDEX will work, and SPACEJUMP
> means ISMAP will work, and there is no way to say that FORM will work.
I think that's because a FORM is an attribute of the document contents,
not the document itself.

> So I am trying to argue here that "SPACEJUMP" and "TEXTSEARCH" are
> really trying to tell you something about the document, not about
> GET per se.
Mostly correct.  It's actually telling you something about the OBJECT in
question.  Not all objects are documents.  However, it also implies
what you can *do* with the object and how it is to be done.  Therefore
it helps defines the interface to the object.

> 3) It is very unlikely that these two keywords convey useful information,
> because most likely, the way you do a GET is by first pulling some
> other document from the server.  Do we really expect it to happen
> that someone will obtain a URL, then do a HEAD on it (to find out
> whether they can use ISMAP syntax to address it), and then do a GET?
No, the browser GET's the URL and says to itself "hey, I can search
using this URL and this method".  Then it changes the user interface
so the user also knows this fact.

> In summary - SPACEJUMP and TEXTSEARCH don't convey useful information,
> have the wrong names for the meanings that they do have, and should
> not be considered methods.  They should be removed.
You are forgetting a very important fact here.  Not all objects returned
by HTTP are HTML.  This information *MUST* be allowed in the object
meta-information (aka the HTTP header) because not all data types allow
ISINDEX and ISMAP!

Perhaps the design should be changed a bit but SPACEJUMP and TEXTSEARCH
should *NOT* be removed in spirit.  I see the names as a pretty arbitrary
since only software ever sees them.

--sanders