Date: Fri, 8 Jul 1994 09:33:27 -0400
From: Jim Davis <davis@DRI.cornell.edu>
Message-Id: <199407081333.AA26477@willow.tc.cornell.edu>
To: ccoprmm@oit.gatech.edu
Subject: Re: New URC Specification is ready....
Michael, congratulations on the URC spec, it is a good step forward.
May I offer a few questions, comments, and suggestions?
1) Perhaps the URC itself requires some meta-data, e.g. the date/time
it was created and the server from which it came. See comment about
TTL below. I am assuming that one might want to keep a URC for
minutes, days, even forever, like a card in a card catalog. In that
case the meta-info becomes more important.
2) I think the precedence rules are a mistake because they are not
flexible. For example, right now, a TTL is associated with the
previous tuple only. But suppose you wanted to add another entry that
was just like the TTL, in that it refered to one and only one entry.
You could not do it, since which ever one came second would not get
the right target. e.g. you want to say
URL: foo
Cost: $25
TTL: 1 day
Bandwidth: low
In your system the TTL would apply to Cost not to the URL. (Hopefully
the indenting makes my intent clear. Indeed I would claim that you
should use indenting, except that you use leading whitespace as a
continuation character.
3) How could you record negative information, e.g. suppose you want
to write down the fact that a certain URL *used* to work, but no
longer does?
4) You need to specify what character set you wish to use and how
extensions to the char set will be encoded, e.g. names in non-ASCII.
5) what if the attribute contains a colon? Is there a way to encode
it?
6) is whitespace after the colon significant or not?
7) Author. How will multiple authors be encoded? How about Corporate
Authors? It would be useful to get a clear standard on this so that
URCs can function as bibliographic records.
8) What is the time origin of TTL? Seconds since when?
9) Is anyone thinking about how to express terms and conditions of
intellectual property usage? I don't think these should necessarily
go into the URC (a pointer to the info would be better.) nor should
the URC spec be delayed while these are designed. But they are
very important issues and *someone* should think about them.
10) Finally, two trivial issues:
1) the HTML document would be easier to read if you used DL instead
of IL. It would also make it easier to automatically find the
attributes, since they'd be tagged with DT.
2) In the example of Abstract, you spelled "laden" wrong.
best wishes
ps to uri-folk - I am not on the uri mailing list but
read it from the archive at lanl.