Re: eventual incorporation of push and hint cache into Squid-HEAD?

Henrik notes:
> > If hint caching is off by default, then while that's true, then the
> > impact is quite minimal.
>
> Define "quite minimal". Is there any reason to why there should be any impact
> at all when the feature is not enabled?

There SHOULDN'T be any impact. But it is 7,000 lines of code. There
may be latent bugs in the interface that don't come out in early testing.

I THINK that each hint cache interface function starts with a test to
check whether hint caching is on. Could be wrong, of course.... I
think you understand what I'm saying.

> > If hint caching is ever on by default, then people not running clouds
> > will see a slightly greater than cache-digest-like loss in
> > performance, and a waste of O(100M) of disk (e.g., a about a percent
> > of the average cache size).
>
> Average cloud cache size I presume? Or the local cache size?

That's for a really large cache cloud. The cache size being compared
to is the local cache, since, just as each proxy cache has its own
cache, each hint cache has its own hint cache.
 
> For a small member in a large cloud the difference is substantial.

Yes. That's true.

But you can't even buy a computer with less than 10-15G of disk these
days.

> Software/algorithm patents are generally a very bad thing in my opinion, more
> so in the twisted direction patent requirements are turning, and will in the
> long run cost human mankind a lot in lost evolution of ideas I
> think.

Amen. I have made it a hobby to look for even one case where any
computer startup has been helped by a patent of any kind while still a
startup. I'm still looking. Every single case has been a big guy
suing, or a little guy suing and not winning. Startups can't afford
lawsuits, end of story.

Hell, startups have being rooked by patents for over a century now.
See the development of the transistor.

> In my opiniton the only valid use of a software/algorithm patent is
> to ensure that it does not get patented by a twisted entity. New
> ideas florish while they are new and interesting, and a patent issue
> is a very good way to deflect the interest from the idea, making it
> starve by forcing the ones not having the patent but interested by
> the idea to find an alternative (and often not as good) solution
> that avoids the patent.

...but our legal system is not able to do that. Litigation is too
expensive for startups to pursue as any kind of strategy. And IP
lawsuits take years to resolve, by which time a computer startup has
either succeeded or not, with or without the burden of competition
from a big copycat, and with or without the burden of legal fees on
the expenses line.

What's more, the correlation between a litigant suing for IP concerns
and 'having a case' is low. Established companies know they can hurt
a startup bad even if they don't have a case, because no judge in the
land understands engineering concerns enough to rule quickly.

> The worst situation is when someone manages to patent something that is in
> widespread use before the patent has been issued, or in some cases even
> filed... The cost/risk in questioning an invalid patent hold by a entity with
> with good fact-bending lawyers is quite substantial, even if the patent
> clearly is invalid to anyone involved.

...but worse still, having an existing patent turns out not to do much
to discourage further patenting of the same idea. All patents claim
to cover practically every idea ever invented, and many as-yet-unknown.

I shouldn't complain so much. Lawyers are a big advance on fighting
things out on the streets.

                                            Jon
Received on Wed Nov 21 2001 - 12:14:15 MST