On Tue, 2006-17-01 at 13:29 +0100, Willem de Bruijn wrote:
> Stackable routes are interesting, and indeed
> simple. They are miles apart from my work, and certainly not innovative. They
> are interesting to this discussion, however.
>
Depends on your definition of innovative - seems you draw a line on
things that havent shown up in academic circles which is weak. Most
people who create these infrastructures in Linux do so because they are
needed not because they eventually want to write a paper so they can get
EU funding[1].
The only paper i know of that sort of mentions stackable dst (this years
after they existed) was in OLS 2-3 years back where they are mentioned
in conjunction with IPSEC.
The paper was from someone in USAGI. Dave also has some examples of
their usage documented on the web (most recent was in chasing an issue
with ipsec SA updates)
I should have pointed the stackable dst myself in my earlier email where
i gave you the two other examples because they are certainly innovative.
It is only fair that you study them before making more comments. Look at
the references given.
> In the discussion of STREAMS vs Linux networking I was not the right person
> to
> give a simple overview. I figured that my response was more useful to John
> than keeping my mouth shut. Bad judgement call, as you guys pointed out
> correctly.
>
Your mistake was to put a plug for your work which you deemed
(implicitly perhaps as) revolutionary ;->
In academic circles there's a lot of hypocrisy - at the end of your
presentation, you get a applause even when people don't find what you
have to offer exciting. On netdev you get honesty in the form of
tomatoes thrown at you.
To give you an analogy: Its like showing up at a pilots-only-vintage-car
show and trying to tell them how you dont ride these vintage cars
anymore, rather you fly like a bird on this thing called a plane which
they should all try out (without knowing these folks are actually
pilots) - what reaction do you expect to get?
Yes, you can get away with this kind of blasphemy in academia.
> I definitely disagree with you here. Linux, Windows and other 'production'
> environments are quite conservative. Research initiatives don't have to be,
> and are therefore often more cutting edge. Compare Inferno, Legion or
> Singularity. Their implementation may suck compared to linux. The ideas
> embedded can form a basis for features in Linux 2.9, however. Hell, 9fs has
> recently been ported.
>
Thats too simplistic an explanation which implies that academia gets it
right. There's also a lot of vaporware and fundware[2] in academia.
- I was reading somewhere that in the days when RED and TCP congestion
control was a hot topic, many papers were written to a buggy ns
implementation. Some of these papers had "ground breaking" thought and
may even have won awards in conferences.
I can assure you that if such ideas showed up in Linux sooner than later
people will figure these out. Thats the way of the real world.
- Also what might look good in academic papers doesnt normally translate
well in the real world. Take an example of all those classification
algorithms showing up every year in sigcomm. Sure, all looks good - but
is classification really our main bottleneck in OS
packet processing? Absolutely not. People wanna solve real world of
dealing with interrupt overload, memory latency, memory copies, context
switches, processor specific optimizations etc. When those 10 issues are
resolved then classification shows up in the list.
So saying Linux (I dont care about other OSes) is conservative is
mistating the issue. The real term is Linux is "real-world".
> > When Van Jacobson did his TCP congestion control work, he learned the
> > BSD TCP stack backwards and forwards before he started making his
> > changes. He made sure to take advantage of what was already there,
> > and add only what was needed.
>
> ofcourse, because his goal was to improve upon this state-of-the-art.
Every research should be on improving state-of-the-art. I guess if i was
to put words in Daves fingers - VJ is a realist.
> Mine isn't.
If VJ is a realist then you are not i suppose - i think thats where the
conflict is for you with everyone else who has responded to you so far.
> I have to learn how these things work, even implement some features
> (such as reassembly) in order to continue with my main work. But nowhere do I
> state that I am an expert on these topics.
>
> That my kernel knowledge is not 100% perhaps disqualifies me as kernel
> hacker.
> This says nothing about my qualifications as researcher, where other
> knowledge is of more importance. Implying that I am not just inadequate at my
> job, but even dishonest ("a charlatan") without knowing what I do is out of
> place
>
You are not gonna get an applause around here. Try to be like the old-VJ
(maybe the new one is still the same as the old) and noone will spank
you ;->
cheers,
jamal
[1] Before you say it is pointing at you - this is just a euphemism to
make a point.
[2] Put something out to get funding
-
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to [EMAIL PROTECTED]
More majordomo info at http://vger.kernel.org/majordomo-info.html
