Re: solidifying a really neat forms hack

Marc Andreessen (marca@ncsa.uiuc.edu)
Wed, 24 Nov 93 06:15:38 -0800


Bert Bos writes:
> [...] I don't think that an ISINDEX element should be equated with a
> FORM. They serve different purposes: the layout and placement of
> ISINDEX is under the control of the browser (a French browser
> displays a French prompt, etc.) and it is *global*, i.e., it is
> available no matter how far you scrolled through the document.

Didn't we just go through this a week or so ago? The behavior you
seem to be claiming is mandated is not in the domain of any spec I've
read. ISINDEX is used to declare that this document represents a
searchable index; it doesn't declare how the search is to be done or
how the user is to enter the search keyword.

> A FORM, on the other hand, is tied to a specific location in the
> text, with suitable prompts provided by the document's author.

Correct -- and what I'm suggesting is that some forms be allowed to
talk to existing ISINDEX-style servers, period. This is independent
of your opinion of how Mosaic 2.0 handles ISINDEX normally. Other
browsers can handle ISINDEX however they want; I'm not trying to
mandate Mosaic's ISINDEX handling method on anyone.

> |So then, Tony Sanders (originator of the incredibly useful ISMAP
> |attribute to IMG -- how is it he comes up with these things before
> |everyone else??? :-) realized that this means that *normal* forms in
> |arbitrary documents that contain, you guessed it, only a single text
> |entry field with NAME="form", provide transparent access to normal
> ^^^^^^
> you mean "isindex" ?

Oops -- yup.

> |I propose that this technique be institutionalized by a one-time-only
> |special case in the forms spec that will insure that all forms-enabled
> |browsers can provide this capability. It's really quite powerful, and
> |[...]
>
> I agree, but it can be independent of the implementation of ISINDEX.
> E.g., specify that the "=" is omitted when the INPUT NAME attribute is
> empty or omitted:
>
> <INPUT name="foo"> ----> some-url?foo=bar
> <INPUT name=""> ----> some-url?bar
> or:
> <INPUT> ----> some-url?bar

Either way a spec modification is needed (since NAME is currently
required for all instances of INPUT). I vote that we go with what
currently works.

Marc