Tim's notes on W5 July 1993

Here are my notes -- someone on the net asked for them. I have separately recorded the wish list for libwww which I came away with. There were many sessions were I didn't take notes. (Who was the official note-taker and where are the results?)

Brainstorming wall

The workshop started with a brainstorming in which we covered the wall with issues which neeed addressing. Here is a dump of the wall. Later some of the issues were tackled by groups of varying (often time-varying!) size. This was a general direction poll, also to help us understand where each other are coming from

Organizational issues

Consortium
Maintenance -- cooperative development
Licensing Software
Net funding policy statements
Design Authority, design pressures
Copyright Issues

BIZ

Business issues

Resource Discovery

Knowbots
Global Indexes, Resource discovery
Notification of changes

HTML

Styles
Multiple DTDs
Whether to talk about PDF etc
Format parameters to negotiate (colours etc) Client profile. See later notes by Tony Sanders .
Forms and User input
Define Implementation leveles fo respecially HTML
Isochronous protocols and sync for multimedia
Discipline-specific descriptions of HTML for certain areas
non-latin charcter sets
Link relationship values
Libwww
Progress feedback -- what does that mean?

Annotation, news

Mail -- to author for example
Context
Versions
Authoring tools, editors,
2400 baud optimisation
"Speculative parallelism" (thanks Simon)

HTTP

Timescales and Impl. levels
z39.50 re;ationships
Authenticaion, Access control and confidentiality.
Semantics of HTTP
Reeverse link building
State
performance

URNS, Naming, URLs

Exec Links
Billing (*action for research)
Caching -- client and gateway (*action for research)
IOnteractive sessions? MuDD?
Reliabiliity and redundancy -- acion for development!
(These notes skip all the group work now)

Last Day

I gave two soapbox microsermons (or was it 3?):

ISMAP: The functionality of specifying that a document is clickable mapstyle should be in the HTTP header not in the HTML of a referring document.
Next and Previous: Browser writers should be aware of different semantics of navigation buttons. General discussion -- awareness, needed, no rules to be laid down here.

MetaInformation

We need to be able to specify the same information either in HTML and/or in a separete HTTP header, as it can logically be put at either level (like email message heading and envelope).

We need a mapping between the RFC822/MIME style and SGML attribute formats in the general case.

Certainly, WWW_Link: is a header which should allow the same information to be specified as the <LINK> tag.

Annotations:

Marc's presentation & general discussions:

Marc Andreessen presented the current annotation system.

Disrtribution methods: Private, Public and group annotations and Newsgroups. Send message around or just a list of things who have new annotations? NNTP is reaching scaling limits..[ Note from Josh Osbourne ]. Retraction might be a useful function.

How to determine the primary publishing name of the document whenit is published in a number of domains?

Finding annotations in edited documents --regular promlem.

How to change HTTP POST for group annotation?

Later small group discussion defined relevenat part of HTTP protocol.

REPLY is no longer a method.
POST and PUT exist.
DELETE added.
POST and PUT have body parts terminated by either length: or Message-Boundary methods, but not CR LF . CR LF method.
GET can always get annotations, from any server. Need reply code saying "I have only got you annotations, not the document".

See current spec which I have updated since.

Conclusions

Personally I found it was great to meet real people behind the mail. Perhaps everyone had a few surprises here, as people's net and real-life profiles are sometime a little different!

I went expecting W5 to be a bit of a zoo, with such a large range of issues, and it was, but it was an interesting and useful zoo. I myself was torn between chairing meetings I was in and jumping in technically with both boots on, wanting always to do both. Next time, we should perhaps make sure chairpeople are appointed specifically to run groups, rather than just suggest one be elected. We will also have enough of an idea of what people want to discuss to prepare an agenda beforehand. Many people (I for one) generally wanted to be in all of the groups at once, though some were quite content to focus one one issue. The total productivity was greater with groups. The HTML+ discussions were like an episode in a long series. There is a lot of work in defining such a DTD.

I got a lot of things out of it. I was glad to see the support for a consortium. I was glad to get lots of feedback on libwww , and offers of help. It was useful to pin down HTTP annotation and front-end update functionality enough or us to do away and implement. And lots of small things cleared up. I am very grateful to Dale for hosting it not to mention feeding us and clothing us!

Tim BL