Re: strategy for HTML spec?

Tim Berners-Lee (timbl@www3.cern.ch)
Thu, 14 Jan 93 10:20:48 +0100


> Date: Wed, 13 Jan 93 14:34:40 CST
> From: Dan Connolly <connolly@pixel.convex.com>
>
> Sounds cool. This points out yet another thing the WAIS folks did
> right: every database has an associated maintainer as part of
> the protocol. WWW could learn a lot from WAIS.

I suggest webmaster aliases for server-wide responsability and
a typed link for authorship of any given document. Many
servers provide stuff from many areas, one person is not
responsible for everything. Our local mainframe doc system
has a function key dedicated to "who is responsible for the runnish
on my screen?!" function. This is important for up-to-date info,
accountability.

I see this as one good use for a link type. Something I have
wondered about a lot is a namespace for people in the URL scheme.
You can even find something about them using SMTP if they
are on the internet.
<A HREF="mail:timbl@info.cern.ch" TYPE="author">tim</A>

Following the link would give options of mailing the person.

> >All we have to do is agree to set up the wwwmaster aliases.

Any objections?

>

> Perhaps HTML2 will provide a handle on how to get ahold of the
relavent
> person -- it should!

You mean http2 or html2?

> >This would also be a good chance to map out the entire Web.
Anyone
> >know any good algorithms for drawing arbitrary directed graphs
with
> >no overlap?
>

> Ever heard of map coloring problems? You can't draw an arbitrary
> directed graph with no overlap. The set of graphs that you can
> draw is the subject of great study, I believe. They're called
> "planar" graphs.

I think we are allowed to let lines cross. Just than minimising the
number of crossings (the "badness") is almose certainly np complete
But in fact a tree will probably be less confusing for people.

I like history trees like

AAAAAAAAAAA
BBBBBBBBBBBBBBB
CCCCCCCCCCCCC CCC C
DDD DDDDDDDD
EEEEEEEEEEEE
AAAAAAAAAAAA *
BBBBBBBBBBBBB *
FFFFFFFFFF FFFFFF
DDDDDD *
GGGGGGGGGGGGGGG

wheer * indicates that the links from that node are listed at a same
or higher level in the tree. (You could give line number
cross-refernece or a link to the listing). This is gives the real
user the comforting illusion that (s)he is playing with a tree, which
people seem to find easier to remember than a web.

> > Which reminds me, a good directed graph is the right way
> >to implement a history mechanism. Somebody want to take a crack
at
> >it?
>

> This is another area where the HyTime folks have done a lot
> of thinking that we should try to take advantage of.

Where do we find the HyTime folk's thinking online?

> Dan
>

Tim