[Date Prev][Date Next][Thread Prev][Thread Next][Author Index][Date Index][Thread Index]
Link Behavior Questions
- To: <tribble>
- Subject: Link Behavior Questions
- From: Mark S. Miller <mark>
- Date: Mon, 16 Oct 89 23:21:47 PDT
- Cc: <marcs>, <xanatech>
- In-reply-to: <Eric>,35 PDT <8910170025.AA01413@xanadu>
Date: Mon, 16 Oct 89 17:25:35 PDT
From: tribble (Eric Dean Tribble)
I'll try to write a full response tomorrow (I don't want to fill my
head with the simulation of your problem today). But a few high level
reactions.
1) let us distinguish two general categories of links: Links
interconnecting documents (which are the type you are considering) and
links internal to documents that represent formatting and the like.
The latter MUST track versions properly. I'm sure a solid case could
be made to blur the distinction between these two categories of links.
Yup. I forgot about these when I replied. I think it is essential
that links encoding text attributes (like font) be embedded links so
they history track correctly. Note that the front-end presentation
problems which are motivating independent links don't apply to these.
Quite the opposite in fact. The front-end presents these links as
integral to text, and it does present embedded text correctly wrt
history tracking.
2) the general class of sliming that you worried about in the latter
parts of your message should be handled by proper filtering. If you
hang a permissive recorder on a public document, you will get a huge
result.
I was about to reply along these lines. I think it is instructive to
go through MarcS's examples and ask whether filtering really is a
sufficient solution for his cases. My intuition is that it isn't. I
think that the link-list displayer would have to notice when multiple
links are actually versions of the same link (because they have the
same link id), and coalesce their display (or some such). See my
response to MarcS.
4) I am slightly terrified about lots of recorders hanging around.
Mostly because I don't really know how expensive Recorders are. The
Sensor part is cheap.
(I suppose the Sensor Canopy is now renamed the Recorder Canopy)
It should be the case (because of the Recorder Canopy) that having
lots and lots of Recorders hanging around is cheap. Log type cheap in
fact. That's what the Recorder Canopy was all about. Am I missing
something here?