We will NOT obey the pricipal that the same reference always accesses the
same hypertext object. We plan a <VAR> tag to indicate variable data in
the text. Text within the <VAR> to </VAR> should not be cached. It would
be helpful if other clients supported this convention. Initially, we will
not cache the entire document is it contains one or more <VAR>, later we
want to be able to request only the <VAR> fields for the document and
merge them with a cached version with minimal communication. VAR should
have a TIMEOUT parameter ultimately.
We will add an <INPUT> tag, with parameters for length, type, and default.
We will allow inputs to be sent using the format of search keywords over
http. It would be helpful if other browsers allow serches on documents
containing the <INPUT> tag. The length suggests how much space to leave
for the input, not a maximum necesarily. Text through </INPUT> may be
the prompt. A name parameter will allow addressing inputs like anchors.
An input is, afterall, like an achor for keyboard data. In addition to
query (search) for input data, we will also require a submit, as http
has no PUT operation, we will use mail (as with comments, see below).
We eventually want an <INCLUDE> tag with an HREF parameter. This will allow
breaking documents into smaller pieces that can be maintained independantly.
A fallback mechanism will be required. I don't know if that issue has
been settled. We will need a rules file capability for the client back
end and support for fallback entries in our rules file. We should consider
restructuring our rules mechanism to better support caching of documents.
We look forward to the dust settling in the multi-media mechanisms for the
web. In the mean time, and in addition, we plan to support file name
conventions to determine the type of the document when it is otherwise
unspecified. Innitially we will support .html, .man, .mm, .ms, .jpeg, .gif,
.au, .tar and the .Z suffix. Support for .mime might also be quite nice
dispite the recursion. We have not resolved all the issues of supporting
this over an http link or dividing the responsibilities between the
client back end and the server. We may innitially restrict this to
local and ftp file addresses.
We have tenativly selected tcl as our machine independant procedure
language to define group activities accociated with the WEB "group
memory". Though we prefer Smalltalk, tcl provides a sufficient meta
language capability across Unix, pc's and mac's in the public domain.
The wish (tk) extensions to tcl for Xwindows and tkWWW are an added
plus. I'd like feedback from others on the use of tcl.
I've been planning an interactive service via http://eies2.njit.edu/
(or telnet://www@eies2.njit.edu/), but lacked time to make it releasable.
I promise to get around to it soon. Comments are sent to www@eies2.njit.edu
as mail with the HREF as the subject and the text of the comment after
a blank line in the message body. If the comment does not begin with
an HTML markup, <FIXED> will be assumed, <> quoted (i.e. \<\>), and anchors
added for imbedded HREFS. (Yes, we do plan to support MIME eventually).
Hopefully the service will be up soon enough for your comments.
Check http://eies2.njit.edu/interactive.
jim