Kendall Clark wrote:
> On Tue, Oct 07, 2003 at 10:16:51PM -0400, ravi wrote:
>
>>> But there is an idea floating around geekdom that the Web works
>>> (in the sense that it scales 5B+ documents, something which no
>>> one really expected) because of various purely technological
>>> ideas...
>>
>> i could use some clarification of the statement above. does it mean
>> that in "geekdom" there is an idea that the web "works" *only*
>> because of technological ideas? if not, then the claim is a truism
>> isnt it?
>
> No, it means that many technical people believe the Web *still* works
> at the present scale because of some specific changes that were made
> to the HTTP protocol. That is, these folks give no credence to the
> alternative explanation that, even w/out those specific technical
> changes, the Web would work at the present scale because of massive
> infrastructural investment...
>
could you point me to some sources? i find it very surprising that
technical people believe that changes to HTTP can be the sole cause of
performance gains (especially given that caching, which indeed does, at
great cost, distribute load, was mostly possible with early HTTP
versions, and further modifications of HTTP were aimed, in a large
sense, at addressing some of the technical defeciencies of a protocol
designed by a non-protocols person, eg: persistent connections).
almost all the technical people i know will readily point to the
increase in network bandwidth (due to the excessive deployment by telcos
in the boom years), the large drop in disk/memory/cpu prices, etc as
significant (perhaps even larger causes) for the gains in scaling. they
would also not find these gain surprising at all. i do not. i do not
find it surprising at all that the internet has scaled to the modest
level it currently is at. i *would* be surprised if we were doing
real-time video over the internet (at the scale of current radio/tv
broadcasting), but that's another beast.
>>
>> 1. what is "the web"? is it the internet + the various web servers
>> and documents that they serve?
>
> That's a good question. I mean in this case it's that part of the
> Internet which happens via HTTP, server & client. It's a significant
> percentage of total Internet usage.
>
so, you are talking about the web component of internet usage? and by
that i assume you mean network usage i.e., available bandwidth and
throughput on routers and other intermediate devices?
>
>> 2. what does scaling to 5b+ documents mean? 5b+ html files stored
>> somewhere on networked computers? 5b+ documents transmitted in
>> parallel (i.e., capacity)?
>
> Neither, actually. There are something like 5B+ addressable resources
> ("things which have URIs"), but they aren't all HTML files on
> servers, many of them are resources which are computed on-the-fly.
> And I doubt anyone believes that all of these resources could be
> simultaneously requested.
>
i use HTML files as shorthand, but if we are talking about dynamic
content, are you including server load and performance? (the previous
point suggests otherwise).
i wouldn't be surprised actually if there comes a day in the near
future, when 5 billion documents are in transit simultaneously on the
internet. i would guess (admittedly a very rough guess) that that number
is already in the millions right now.
and as more and more people blindly adopt HTTP as their transport
protocol (simply because of such technologies as "web services"), often
ignoring the years of work done with protocol design, many of these
documents will be transported over HTTP (though it might not be the
right transport at all) -- to take the rant a bit further about the
ascendancy of buzzword compliance since the "corporatization" of the
net and the IETF, the current craze with XML was well ridiculed by a
recent RFC (an internet technical specification document) which parodied
this trend by putting forward a proposal for "IP over XML"!!
http://www.faqs.org/rfcs/rfc3252.html
>>
>>> what does "default libertarian geek mind" mean? that by default
>>> you assume all geeks are libertarians? or that you have found
>>> them to be so?
>
> I mean that the dominant ideology among the geek set (well, large
> chunks of it anyway, it's probably not more monolithic than any other
> subculture) is strong right libertarian, especially on the issue of
> where technology comes from. It's *not* a David Noble-friendly part
> of the world, at least as I have experienced it. (And, yes, I do tend
> to assume that most geeks are right libertarians, given the dominant
> ideology, but it's a loose assumption which I stand ready to modify.
> Anyway, not sure how this is relevant...)
>
i am not sure how this is relevant either, but hey, i didnt mention it
;-). you must have thought it relevant, otherwise why would you mention
it? ;-) and as a geek, of course i take offense! seriously however, all
the geeks i know are somewhat of a mix of humanitarian or analytical
leftist. of course we might differ on what we consider a "geek". perhaps
this is a west vs east coast thing?
the IETF (or perhaps the IAB or IESG, i forget who authored the
document) for instance suggests that it is neither a dictatorship nor a
democracy, but that it works by technical consensus (if you believe
some) or as a meritocracy (in the words of others). in the words of dave
clark: we reject kings, presidents, and voting -- we believe in running
code! would you call that a libertarian viewpoint?
>
>>>>> Sorry, but I wouldn't dream of asking an actual computer
>>>>> technical question on PEL-L or LBO. :>
>>>
>>> why not?
>
> Because it's completely off-topic? Isn't that obvious?
>
its obvious that its off-topic, but its not obvious (at least to me)
that that's why you wrote the above. michael has been quite lenient
towards computer tech questions on this list and people have asked them,
and some have even got answers!
> I've already explained it, so I won't do so again. I'm not gonna go
> 'round and 'round about this, Ravi, since it's not really germane to
> my question. I'm starting to regret including any surrounding
> context.
you have to realize that i ask these questions because:
1. what you specified as the context was not clear to me. it still is
not (and probably because i am not reading you right).
2. i am surprised by your generalizations about the geek and computer
science community. perhaps what you mean by geek is the high-school geek
set while what i mean is the hacker crowd (for the general audience:
'hacker' does not mean what the media has wrongly used the term to
represent i.e., someone who breaks into computers). i have lived among
the hacker and computer science community for 15 years now (including a
long stint at one of the temples: bell labs) and your statements do not
match my experiences very well. if that is because i have misunderstood
my community, then i would appreciate any clarifications that disabuse me.
while these might be peripheral to your main question, once you put
these opinions out in a public venue, i think discussion on them is
valid. of course, if michael thinks we should go off-list, i will gladly
do so.
--ravi
