You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@httpd.apache.org by Henrik Frystyk Nielsen <fr...@w3.org> on 1998/02/10 18:58:53 UTC
Re: W3C <-> Apache
At 10:32 2/9/98 -0500, Daniel Veillard wrote:
>> > >Any rough timelines on it?
>> >
>> > Nope.
>>
>> Yep June 97, at that time we will give an answer to the initial question
>> whether an object-oriented model is suitable for HTTP and will allow a
better
>> evolution of the protocol. Maybe no, maybe yes, in both case we will
>> justify our answer with more than words ...
>
>Right. And you'll have done it without _any_ consultation with the
>developers of the most widely used HTTP server. Good plan! (not).
Hi Ben,
I was forwarded your note from Daniel Veillard - I am not subscribed to the
Apache developers list, so please forgive me as I haven't seen the rest of
the thread.
There are some often seen misconceptions in your mail that I want to put
straight:
1) Web experts like you folks are most welcome to participate in W3C work
and it is not a general requirement that you are from a W3C member. What
_is_ a requirement is that when you sign up, you donate a significant part
of your time to actually work on the project. This rule is in order to make
sure that we don't have a lot with good intentions but little time to do
much about then other than producing a lot of mail.
It is also a requirement that you follow the rules of the chair of the
working group of which parts of the project is public and which is member
only. The reason for this is that the W3C is now in a position where the
press follows everything we do and say and when and if things leak, it
takes a lot of time to put things right.
That is why I am very concerned about the recent leaks of information onto
this mailing list that I rightfully understand have made you upset. The
information that I have seen go by did not represent HTTP-NG in a rightful
way. I would be happy to help in any way that I can but it requires that
somebody from the Apache group signs up to participate.
2) W3C is committed to work with the Web community following the rules of
running code and rough consensus. W3C has been a first class player in all
IETF WGs where we have participated, joining in on both freely available
sample code and spec writing. I think that the development of HTTP/1.1
pipelining was an excellent example of how I worked together with Dean
Gaudet on improving both the libwww sample code and the Apache HTTP/1.1
support.
3) We are not in any way trying to define a new HTTP protocol without the
collaboration with IETF, Apache, or any other significant player in the Web
community. This is a proto type project where we try to find out if we can
base the current Web model on top of a distributed object system and still
maintain the flexibility and simplicity of the Web. I have many good
reasons for why this might be a good idea but it takes time to actually be
able to evaluate it in practice.
In HTTP-NG, we have been looking into how we can use the Apache server for
handling HTTP-NG - exactly for the reasons you mention - it is a great
server with a great performance! We have not been able to get Jigsaw to
perform as well due to its Java based implementation.
But let me make it clear that as with everything in life, you have to ask
to get on board. In the case of HTTP-NG we ask for 50% of your time, I know
it's a lot but it is helping us guarantee that we actually make progress.
Again, I would be very happy to see active participation from the Apache
group - until now I haven't heard such a request - but would be happy to
get input from 1-2 people!
If you do not think that 50% is a reasonable or practical amount of
participation then we have another forum called the HTTP-NG Interest Group
where participation isn't constrained to participation but of course the
impact on the project is lower.
Thanks,
Henrik
--
Henrik Frystyk Nielsen,
World Wide Web Consortium
http://www.w3.org/People/Frystyk
Re: W3C <-> Apache
Posted by Rodent of Unusual Size <Ke...@Golux.Com>.
Henrik Frystyk Nielsen wrote:
>
> I suggest that instead of throwing a lot of big (and largely meaningless)
> words at each other, I think we should work on how to proceed. This
> requires, however, that we start using a more constructive tone instead of
> calling each other for shit heads.
Sounds good to me. Remember that's how you should interpret my
remarks below..
> As I said, I would very much like to get more input from the Apache group -
> I do need people who can do real work. I have made the conditions clear so
> here you go - it's up to you!
Maybe it isn't obvious, and maybe it's just my opinion - but I should
think that the group responsible for designing HTTP-NG would do well
to *solicit* input from the group responsible for producing the most
popular Web server on the planet. Instead, we sort of get into this
discussion by accident, and what I'm seeing could easily be construed
as "here are the rules: give us 50% of your time or don't expect us
to take you seriously."
That doesn't sound like a collaboration; it sounds positively
adversarial and elitist. I doubt that's the intent at all, but I
wanted you to know how it could be interpreted.
I think it incredibly unlikely (less than 0.05 probability) that anyone
in The Apache Group can dedicate 50% of their time to HTTP-NG. Is there
any hope of equivalency? You know, "BsC *or* 5 years experience
required"? We have the experience (well, as a group); can't that
be taken in lieu of dedicated time in terms of measuring impact?
#ken P-)}
Re: W3C <-> Apache
Posted by Rodent of Unusual Size <Ke...@Golux.Com>.
Ben Laurie wrote:
>
> Henrik Frystyk Nielsen wrote:
> >
> > Since when has W3C been in any market? If you read our contract then we
> > explicitly state that we don't compete with anyone.
>
> Oh come on! W3C is a consortium. Of who? Of Web developers. W3C may
> claim to not compete, but its members can't make that claim. I love this
> idea that you can state in your "contract" (who is this a contract
> with?) that you don't compete, and somehow, magically, its true.
Come, come - let's try to work together towards a means of
collaboration and extending mutual trust and respect. Enough
self-righteousness - on both sides.
Henryk, I think several of the Apache developers would like to
be involved in HTTP-NG design/evolution. However, they probably
can none of them dedicate 50% of their time. Will W3C cut them some
slack? Is 50% of someone's time who can barely spell HTTP worth
more to W3C than 5% of Roy's, Marc's, Jim's, Dean's, ... time?
#ken P-)}
Re: W3C <-> Apache
Posted by Ben Laurie <be...@algroup.co.uk>.
Henrik Frystyk Nielsen wrote:
> >Now, in case people think I'm completely serious about this: of course,
> >I'm not. I'm more reasonable than W3C. OTOH, the temptation of 50k a hit
> >could make me much less reasonable. But seriously: in what way is what I
> >have said different to W3C's approach? If anything it should carry more
> >weight. We represent more market than they do, after all.
>
> Since when has W3C been in any market? If you read our contract then we
> explicitly state that we don't compete with anyone.
Oh come on! W3C is a consortium. Of who? Of Web developers. W3C may
claim to not compete, but its members can't make that claim. I love this
idea that you can state in your "contract" (who is this a contract
with?) that you don't compete, and somehow, magically, its true.
Cheers,
Ben [who only ever tells the truth. It says so in his "contract"].
--
Ben Laurie |Phone: +44 (181) 735 0686|Apache Group member
Freelance Consultant |Fax: +44 (181) 735 0689|http://www.apache.org
and Technical Director|Email: ben@algroup.co.uk |Apache-SSL author
A.L. Digital Ltd, |http://www.algroup.co.uk/Apache-SSL
London, England. |"Apache: TDG" http://www.ora.com/catalog/apache
Re: W3C <-> Apache
Posted by Henrik Frystyk Nielsen <fr...@w3.org>.
At 23:28 2/10/98 +0000, Ben Laurie wrote:
>Amazing! You think you can dictate conditions to us. As the de facto
>setter of standards for Webservers, I say that you must follow our
>process, namely to submit patches to Apache and gain appropriate support
>for those patches. Until you do, you have not followed the due process
>for changing the way the Web works.
That would work but I was actually proposing a mechanism for working closer
together. I guess your proposal would make that harder. I don't
particularly think that is a good idea and it is not what I have heard
mentioned in today's discussion.
>Now, in case people think I'm completely serious about this: of course,
>I'm not. I'm more reasonable than W3C. OTOH, the temptation of 50k a hit
>could make me much less reasonable. But seriously: in what way is what I
>have said different to W3C's approach? If anything it should carry more
>weight. We represent more market than they do, after all.
Since when has W3C been in any market? If you read our contract then we
explicitly state that we don't compete with anyone.
Henrik
--
Henrik Frystyk Nielsen,
World Wide Web Consortium
http://www.w3.org/People/Frystyk
Re: W3C <-> Apache
Posted by Ben Laurie <be...@algroup.co.uk>.
It's late, and I have things to say about much of this tomorrow, but I
can't let this one go...
Henrik Frystyk Nielsen wrote:
> As I said, I would very much like to get more input from the Apache group -
> I do need people who can do real work. I have made the conditions clear so
> here you go - it's up to you!
Amazing! You think you can dictate conditions to us. As the de facto
setter of standards for Webservers, I say that you must follow our
process, namely to submit patches to Apache and gain appropriate support
for those patches. Until you do, you have not followed the due process
for changing the way the Web works.
Now, in case people think I'm completely serious about this: of course,
I'm not. I'm more reasonable than W3C. OTOH, the temptation of 50k a hit
could make me much less reasonable. But seriously: in what way is what I
have said different to W3C's approach? If anything it should carry more
weight. We represent more market than they do, after all.
Cheers,
Ben.
--
Ben Laurie |Phone: +44 (181) 735 0686|Apache Group member
Freelance Consultant |Fax: +44 (181) 735 0689|http://www.apache.org
and Technical Director|Email: ben@algroup.co.uk |Apache-SSL author
A.L. Digital Ltd, |http://www.algroup.co.uk/Apache-SSL
London, England. |"Apache: TDG" http://www.ora.com/catalog/apache
Re: W3C <-> Apache
Posted by Marc Slemko <ma...@worldgate.com>.
On Tue, 10 Feb 1998, Henrik Frystyk Nielsen wrote:
> At 12:57 2/10/98 -0700, Marc Slemko wrote:
> >On Tue, 10 Feb 1998, Henrik Frystyk Nielsen wrote:
> >
> >[...]
> >> That is why I am very concerned about the recent leaks of information onto
> >> this mailing list that I rightfully understand have made you upset. The
> >> information that I have seen go by did not represent HTTP-NG in a rightful
> >> way. I would be happy to help in any way that I can but it requires that
> >> somebody from the Apache group signs up to participate.
> >
> >While the gooals may be fine, the method has serious flaws from where I
> >sit.
>
> Marc,
>
> I suggest that instead of throwing a lot of big (and largely meaningless)
> words at each other, I think we should work on how to proceed. This
> requires, however, that we start using a more constructive tone instead of
> calling each other for shit heads.
I am not trying to insult anyone, but am simply saying how things look
from here and why I am unhappy with them. If it did not come across
as intended, I apologize. I am sometimes... erm... abrupt in words.
[...]
>
> - The Web Characterization Group analyses existing log files to
> find patterns and distributions (requires 50%)
> - The Protocol Design Group tries to design a protocol that can
> fulfill the needs put forward by the WCG (requires 50%)
> - The HTTP-NG Interest Group is for interested parties that
> don't have the time/resources to participate 50% but want
> to follow discussions on a mailing list. Of course, the
> level of influence is set accordingly.
I wasn't able to find any information about any of the above anywhere.
Could you give me a pointer to a URL where they are discussed? In
particular, the HTTP-NG Interest Group and exactly what is required to
"join" and if it is possible to simply monitor what is going on to ensure
that the vision for HTTP-NG fits in with my vision for Apache.
My main interest in HTTP-NG is from the network viewpoint and ensuring
that the appropriate issues are taken into consideration WRT network
traffic and interactions and also from the perspective of seeing just what
sort of idealogical view is required in a server to support it properly.
The fact that it is not yet beyond loose talk or whatever point it may be
at is irrelevant. By the time it is, it will be too late for anything but
relatively minor corrections.
Re: W3C <-> Apache
Posted by Henrik Frystyk Nielsen <fr...@w3.org>.
At 12:57 2/10/98 -0700, Marc Slemko wrote:
>On Tue, 10 Feb 1998, Henrik Frystyk Nielsen wrote:
>
>[...]
>> That is why I am very concerned about the recent leaks of information onto
>> this mailing list that I rightfully understand have made you upset. The
>> information that I have seen go by did not represent HTTP-NG in a rightful
>> way. I would be happy to help in any way that I can but it requires that
>> somebody from the Apache group signs up to participate.
>
>While the gooals may be fine, the method has serious flaws from where I
>sit.
Marc,
I suggest that instead of throwing a lot of big (and largely meaningless)
words at each other, I think we should work on how to proceed. This
requires, however, that we start using a more constructive tone instead of
calling each other for shit heads.
To explain where we are coming from, it's clear that the Web will (if not
already) have merged with some sort of a distributed object system. Until
now, the most serious attempts have been to base it on top of HTTP using
POST. While this is a viable way to go, it is not likely that it will be
the best way to ensure the evolution of the Web.
On the other hand, neither of the distributed object systems around have
any of the characteristics of a scalable, lightweight system useable on the
Internet at large.
The question that we are trying to solve is to see whether it is possible
to make a distributed object system that actually meets the requirements of
the Web. Now what are these requirements?
That is not quite so easy to see which is the reason why we have two
Working Groups as part of the HTTP-NG Project:
- The Web Characterization Group analyses existing log files to
find patterns and distributions (requires 50%)
- The Protocol Design Group tries to design a protocol that can
fulfill the needs put forward by the WCG (requires 50%)
- The HTTP-NG Interest Group is for interested parties that
don't have the time/resources to participate 50% but want
to follow discussions on a mailing list. Of course, the
level of influence is set accordingly.
There are many people who have ideas about how and if a distributed object
can be used as a foundation of the Web. We are actually trying to make a
testbed where we can run scenarios provided by the WCG to see how things
work out.
I actually talked about building a testbed, making rough implementations,
and taking data which is why we haven't used a lot of time on writing
documentation. Our experience from HTTP/1.1 shows that this is the only way
to guarantee that a design works - we don't want to be in the same
situation that Tim, Roy, I and others were in when HTTP/1.0 was designed.
>> If you do not think that 50% is a reasonable or practical amount of
>> participation then we have another forum called the HTTP-NG Interest Group
>> where participation isn't constrained to participation but of course the
>> impact on the project is lower.
>
>Erm... I'm afraid I must be missing wherever or whatever this group may
>be.
As I said, I would very much like to get more input from the Apache group -
I do need people who can do real work. I have made the conditions clear so
here you go - it's up to you!
>AFAIK, the current Apache core has no hope in hell of implementing
>presumed parts of HTTP-NG like MUX without serious changes to the process
>model. While these changes are planned anyway, there can be no conceptual
>planning to be sure the two can match up right--and avoiding yet another
>major set of changes when HTTP-NG does magically appear.
Right now, it doesn't make sense to talk about large scale deployment - we
don't know how it is going to turn out. The HTTP-NG project has until June
to try and come up with enough arguments to make a better judgement and to
be able to demo parts of the concepts.
Henrik
--
Henrik Frystyk Nielsen,
World Wide Web Consortium
http://www.w3.org/People/Frystyk
Re: W3C <-> Apache
Posted by Rodent of Unusual Size <Ke...@Golux.Com>.
Henrik Frystyk Nielsen wrote:
>
> At 11:25 2/10/98 -0800, Roy T. Fielding wrote:
>
> >The general problem is that "The Apache Group" operates by a principle
> >of public disclosure of every technical discussion and decision ever made.
>
> That is a bold statement - all W3C results are indeed also made public -
> otherwise we would be sued on anti-trust grounds. But don't tell me that
> all discussions are taken based on public input and consensus.
These are not the same thing at all. Roy is talking about public discussion,
while you mention public results. One puts the public into a contributory
position, in front of the power curve as it were, and the other treats
the public strictly as a consumer.
Neither the AG nor the W3C operations are so black and white in process,
but I think the distinction is worth making. I'm not passing any
sort of judgment on either model.
#ken P-)}
Re: W3C <-> Apache
Posted by Henrik Frystyk Nielsen <fr...@w3.org>.
At 11:25 2/10/98 -0800, Roy T. Fielding wrote:
>>That is why I am very concerned about the recent leaks of information onto
>>this mailing list that I rightfully understand have made you upset. The
>>information that I have seen go by did not represent HTTP-NG in a rightful
>>way. I would be happy to help in any way that I can but it requires that
>>somebody from the Apache group signs up to participate.
>
>You need to be clear on the difference between an opinion and a leak.
>W3C doesn't want any of the work products to be published until they
>have at least a minimum of approval from members.
Roy,
This is actually not true - a Working Group can publish a Working Draft to
the public or to the membership without member approval. It's only when a
document is about to become a recommendation that the Membership is asked
to vote on it.
> I didn't include
>any of those products in my messages, and was careful to describe my
>opinion of what the HTTP-NG PDG is doing rather than including any of the
>actual specs/code that you guys have produced so far. That is no different
>than what has already been published by the W3C. I was even reticent to
>reveal that Daniel had done some coding work with the Apache/ILU stuff,
>but that represents his "player ID card" in a group where developers get
>more respect than non-developers, and he'll need that card to support
>any comments he has on the 2.0 redesign.
>
>As for "did not represent HTTP-NG in a rightful way", I can only say that
>my summary was as close to accurate as possible without revealing any of
>the technical details. [Keep in mind that I've spent most of the past
>six months researching the advantages/disadvantages of distributed object
>systems versus state representation systems (HTTP), so my summary is
>probably a lot more accurate than your own documents.] To do better,
>you will have to make the technical details public so that they can be
>considered independent of other people's opinion.
I think that sounds very interesting and I would love to hear more about
this. We have actually not used a lot of time on the documents as we have
tried to get the testbed up and running. I hope that we have time to write
some more documentation this spring.
>The general problem is that "The Apache Group" operates by a principle
>of public disclosure of every technical discussion and decision ever made.
That is a bold statement - all W3C results are indeed also made public -
otherwise we would be sued on anti-trust grounds. But don't tell me that
all discussions are taken based on public input and consensus.
>That is a polar opposite to the notion of a member-supported consortium
>where the consortium activities are performed in secret.
No, our activities are not secret - you can go to our public web site and
find information about all our activities. You may not be able to find
everything about the activities but I do think that is pretty normal, even
within the Apache group, or so I hear ;)
> Ben's problem
>is with how W3C operates, and that opinion is shared by most free software
>developers. So, while I think you've done a pretty good job in inviting
>experts to join, it is still not an open process. That doesn't mean it
>needs to be an open process (design by committee is not fun), but you
>will have to expect that people outside the process will treat it as
>suspicious, at least until you give them enough clues to understand
>what it is you are actually working on.
It would be the same if everybody knew everything about your research
projects - it is not possible nor desirable and it will slow things down.
The same is the case of IRTF and the reality of many IETF WGs as well.
It's all a question of timing. It has been made very clear that when we
have something ready we will go to the IETF and ask for the appropriate set
of Working Groups to be created and continue the process there. We are
currently in the exact same position as you are - we are investigating how
and if this may work and we are actually trying to implement our ideas.
Henrik
--
Henrik Frystyk Nielsen,
World Wide Web Consortium
http://www.w3.org/People/Frystyk
Re: W3C <-> Apache
Posted by Dean Gaudet <dg...@arctic.org>.
On Tue, 10 Feb 1998, Marc Slemko wrote:
> AFAIK, the current Apache core has no hope in hell of implementing
> presumed parts of HTTP-NG like MUX without serious changes to the process
> model. While these changes are planned anyway, there can be no conceptual
> planning to be sure the two can match up right--and avoiding yet another
> major set of changes when HTTP-NG does magically appear.
I can't emphasize this point more. We're about to redesign. I'd like to
know now if our redesign is going to let us do http-ng easily. It'd be a
waste to redesign twice.
None of us have 50% of our time to spend on Apache, let alone on HTTP-NG.
Dean
Re: W3C <-> Apache
Posted by Marc Slemko <ma...@worldgate.com>.
On Tue, 10 Feb 1998, Henrik Frystyk Nielsen wrote:
[...]
> That is why I am very concerned about the recent leaks of information onto
> this mailing list that I rightfully understand have made you upset. The
> information that I have seen go by did not represent HTTP-NG in a rightful
> way. I would be happy to help in any way that I can but it requires that
> somebody from the Apache group signs up to participate.
While the gooals may be fine, the method has serious flaws from where I
sit.
By saying that anyone who has any access to know anything about the
protocol (even knowning a couple of words about what it does and what sort
of rough timelines are being considered appears to be forbidden knowledge)
needs to devote most of their time to it you ensure that you have no one
involved in any way who is currently firmly planted in the real world of
the here and now of the web. No insult intended, but however much
background you have you end up with a very different view when you devote
most of your time to thinking up new schemes instead of spending most of
your time trying to hack together the existing ones to work. You have no
way for the people who are implementing protocols right now and who will
have to implement this new protocol if it is released to have any concept
of what will happen. There is no room for early feedback from
implementors and no room for those implementors to be sure they are moving
in a general direction to be able to have the framework to implement the
protocol.
While I fully understand the necessity of keeping the number of people
involved at a minimum and to have them actually be constructive in order
to progress at a reasonable speed, you can only go so far before you
progress at a wonderful speed on something that is useless.
[...]
> If you do not think that 50% is a reasonable or practical amount of
> participation then we have another forum called the HTTP-NG Interest Group
> where participation isn't constrained to participation but of course the
> impact on the project is lower.
Erm... I'm afraid I must be missing wherever or whatever this group may
be.
Right now, HTTP sucks. Badly. This is no insult to the people who put a
lot of good effort into making it what it is today, but simply a
reflection of the fact that it has gone a lot further than it was
originally designed for. There are some serious changes that are required
to ensure it can work efficiently and cooperate well with networks in
addition to filling the needs of the user. If there is no observable
movement in that directly with HTTP-NG (and, from the perspective of
anyone who can't devote their life to it, there isn't), you will end up
with people trying to do their own thing. That does no one any good.
I don't know what stage the HTTP-NG work is at, I don't know if there is
anything that could be made known about it yet, I don't know what the
plans are for doing that, but I do know that it doesn't look very
promising from where I sit, and my network isn't happy at the thought of
having to keep digesting HTTP/1.[01] traffic forever. It is running out
of tums.
AFAIK, the current Apache core has no hope in hell of implementing
presumed parts of HTTP-NG like MUX without serious changes to the process
model. While these changes are planned anyway, there can be no conceptual
planning to be sure the two can match up right--and avoiding yet another
major set of changes when HTTP-NG does magically appear.