Date: Fri, 26 Nov 93 13:38:44 +0100
From: Tim Berners-Lee <timbl@www3.cern.ch>
Message-Id: <9311261238.AA00233@www3.cern.ch>
To: Peter Deutsch <peterd@bunyip.com>
Subject: Re: What we're trying to do here...
Peter Deutsch <peterd@bunyip.com>
uri@bunyip.com
Peter, you have misunderstood most dramatically and I hope
not intentionally.
>From: Peter Deutsch <peterd@bunyip.com>
>Date: Thu, 25 Nov 1993 16:13:57 -0500
>
>[ Tim wrote: ]
>. . .
>> It is this fearsome urge to create taxonomies to constrain things
>> which are different into boxes of nominal similarity which is
>> preventing progress.
>. . .
>[ and later in the same message wrote: ]
>
>> We (the URI WG) are not inventing things with the URL stuff.
>> We are documenting existing practice. The inventions go into the
new
>> schemes, not the old.
Peter:
>Given our known disagreements in the past over this subject
>I've tried to stay out of the current debate over the URL doc,
>but I'm afraid I can't let this shot go past without at least
>some comment. Despite my reputation, I'll try to be brief...
>
>
>I think your remarks here illustrate very well where the
>source of problems over the URL doc have been in the past 18
>months. Given your apparent belief that a standards setting
>working group is tasked only with documenting existing
>practice ...
I'm sorry that was NOT my belief. I said "with the URL
stuff" and if I had meant it in general I wouldn't have said
"with the URL stuff".
I think the WG should
be inventing things and not trying to classify RFC850
which I call existing practice.
I am saying that when it comes to the *URL* document
we are documenting existing namespaces, whose propoerties
are defined by the relevant protocols (usenet, nntp, ftp, etc).
I am *not* saying that the the or any working group should
only document existing practice.
> ... and your apparent rejection of the other components
>of the IIIR architecture (in particular, your apparently
>problem with the concept of URNs),
Not at all. I think Mitra and I are the only people to actually
suggest a way making URNs work.
> your resistance over the
>past year to incorporating into the working group document
>changes to your own version of a URL spec is understandable.
I am sorry - as I only put changes in when they seem to be agreed.
I am putting those things which were agreed, but most points
were "postponed to the list".
>Unfortunately, I believe you to be wrong on this issue for at
>least two reasons.
Which issue? You have compleetly misinterpreted what I believe...
>First, there are a number of other people besides the WWW
>community who have been working with the URI working group to
Of course. And there are more people than bunyip. Give us
a break.
>...ensure an orderly transition to the proposed new world of URNs
>and URLs.
URLs describe existing protocols. You want to make the existing
world more orderly by declaring it to be different. Like the
guys who want to tax fruit and vegetables differently, you don't
know what to do with the tomato. Declare it as officially
a vegetable though biologically a fruit? Or only allow it
in desert menus because it really is a fruit? Change what it is
or how it is used? Oh dear, the URN/URL scheme is such a nice
invention, but what shall we do with the nntp tomato? It is
too much like a URN: let's change it to be a real URL.
(Let's just eat the roots of the tomato as a vegetable)
It's really a URN: let's prohibit its use as a URI because
it isn't a URL and we haven't invented URNs yet.
(And besides, should it be fitted nasally? -- D.Adams)
> ...To now argue that we can _only_ consider
> existing practice (that is, the WWW URL scheme)
This is flagrant misrepresentation. I am NOT talking about WWW.
The existing use we are talking
about is for example NNTP. Look at eth "References": header in
a news message. There is a reference. What do you see?
Is it a group/number identifier? All WWW did was to map what is
provided by the protocols. The fact is that news articles
DO ALREADY quote message ids (inthe body of the text often too).
FTP filenames are quoted without specifying the transfer mode.
This is the world as it is, and was pre-WWW. Sure we want to
make a better new world, but we can't retrospectively redesign the
tomato.
>Second, as others have informed you in the past it is _not_
>the purpose of IETF standards setting working groups to merely
>document existing practice when creating a standard.
I never said it was.
> In fact,
>I would argue that such an approach would not be engineering
>at all, but merely documentation and the work of informational
>RFCs at best.
In this case, the RFCs for network news
have already been published. The NNTP WG should be
revived if the way of referring to a message will in
future be by host/article
>In particular, we must keep in mind that URLs are certainly not
>the last word on integration of Internet information services.
>In fact, they are only the first, and their design should take
>into consideration the needs of future components of this new
>system, even where some changes to existing practice are
>therefore required.
>Those of us who have argued the need for a functional
>specification to guide our work, for prefix ID strings, for
>reworked news URLs, and so on have all said that our own work
>would be better served if these were in place.
>You are free to object when proposed features impact your
>installed user base
I am sorry if you feel that by having an installed base
the WWW community have less feel for what in the long term will
work and what will not. That is all which concerns
me. The WWW community has made changes before and will
make changes again when it comes to maingthings work which
don't work.
I am simply worried that the URI WG is making a broken design.
I am worried that the IIIR world will be screwed up for everyone.
My worries:
The world will be full of references of the nntp:
type which people just won't be able to follow, and
the whole system will break.
I am worried that it will remnain broken, while
the IIIR group comes up with "the definitive"
URN protocol, and then people won't be able to read news
without the URN protocol being up to.
I am worried that that will exclude most usenet people.
I am worried that by demonstrating that the integrated
system can't even include the existing usenet world,
we will suspect that it won't be able to include the
next, rather different protcol tomato which is better.
>In short, as the leader of the only Internet service project
>which currently deploys a form of URL you seem to have been
>labouring under the misapprehension that what we were all
>doing was meeting to bless your work as a new Internet
>standard for others to follow.
Sorry, this is not true of URLs. It was true of HTML which
should be released as informational.
There is quite a strong element of personal attack in your message.
I imagined that you behaviour in Amsterdam followed from the
strain of starting a company, and as you did not turn up to the
following meeting I was unable to discover any better.
> Far from it. The way I see it
>we have all spent the last 18 months creating a new integrated
>information architecture for the Internet and drawing upon the
>work of many people, yourself included, in the process.
And what have we come up with? We have not managed to
fix on URLs, when we should be implementing URNs.
>We now have a URL format that reflects the needs and desires
>of a wide selection of Internet developers and user
>communities (although I doubt if anyone is 100 percent happy
>with the current proposal).
I wish I could believe that it does reflect that to a greater
extent than previous deafts. But I will put it into the
document anyway.
>I hope we can, at this late date, persuade you to accept a
>change in your world view and endorse both the process and the
>result. It would be nice if your misunderstanding of what
>we've accomplished be worked out before the document be
>advanced onto the standards track, although I would also point
>out that despite your success with WWW you do not have a veto
>on this group's output.
What was demonstrated in Amsterdam, Peter, was that despite
a strong commercial interest, a product still under wraps
with no published specifications, you *DO* have a veto
on the group's output!
I really do hope this misunderstanding can be worked out.
If, as Eric said, we didn't have consenses then, we don't
seem to have it now. For example, the URL: prefix was
decided at Amsterdam quite decicively I thought
(that it would not be in) and in Houston it was by
rough consensus (I heard noise in both directions but the camera did
not get the hands) it was put back in. No -- that is what I heard
on a mail, but the minutes said it was up for discussion.
>Others have accepted things they disagreed with to accommodate
>you in this work and I hope you can see the value of
>accommodating the needs of others here, too.
I hope that many of the changes which
have been introduced assure you that I have been happy to change.
Character sets, for example -- having suggested strongly that
spaces should not be allowed
I accepted their reintroduction -- largely following from your
proposal that a completely binary format be used -- and now
I see I have to put them back in.)
As there is clearly a call for it, I have put NNTP: URLs
in with a warning note (don't panic: news URLs have TWO warning
notes!).
> - peterd
There was an element of personal attack in your message.
I hope that this was based on misunderstanding.
I certainly try at all times to be constructive, open and fair.
I shall also continue to shout "tripe!" whenever I believe tripe
to be served.
Tim BL
CERN