Re: [9fans] (no subject)

[Roman Shaposhnik <rvs@sun.com>]

On Dec 6, 2008, at 11:28 AM, Dave Eckhardt wrote:
> More globally, if the high adoption rate of NFS is an argument
> in favor of its architecture,

It is most definitely not. At least in my opinion. However, adoption
is the only thing that I know of that can potentially justify excessive
*engineering* complexity. Personally I'd call any system that is overly
complex *and* has a low adoption rate (for whatever reason) after
at least 5 years of having been exposed to the market place, a
theoretical artifact of computer science. The usefulness of which
is commonly measured by the number of good CS papers dedicated
to studying it.

I do understand that for you AFS also has a huge redeeming factor
of being useful in solving a particular problem. It is hard to say
whether
it is the best tool for the job, or the only one (it would be
interesting if
you answered to Erik's question on this list). But perhaps the
difference
of opinion here stems from the fact that to *you* it does feel like
an "enterprise grade" to me (as an outside observer) the "enterprise
grade" must be vetted by the market place. Hence my interest in
the adoption rate.

> and the low adoption rate of AFS is an argument against its
> architecture,
> why are you reading a Plan 9 mailing list...?

This one is easy: Plan 9 (and 9P in particular) doesn't have to have
the redeeming quality of high adoption rate in order to justify
an excessive engineering complexity. It is not complex at all.
It is small and elegant. Whether that compactness and elegance
sometimes prevents it from being considered "enterprise grade"
is an open question (at least for me it is). The experience of
Coraid suggests that it might actually be a nice tool even for those
kinds of problems.

>> To some extent, the popularity of NFS (is there any NAS box that
>> talks AFS?) and Linux is one big testament to the power of "good
>> enough" or "worse is better".
>>
>> Designing "enterprise grade" things is very hard work.
>> Implementing them is even harder. The good news is that it
>> pays well. The bad news is that you have to be really brave to
>> withstand the fear of being obsolete by changing requirements.
>
> I don't get this.  I don't follow the NFS protocol development
> carefully,

Just to clarify: this thread is really not about me defending NFS.
I can't call it elegant. But I can certainly call it "good enough".
So at least it succeeds as a lowest common denominator for
data sharing. Well, so far at least.

> That is, I think the requirements are *not* changing, but rather
> that NFS is slowly realizing that those things *are* requirements.

Agreed. And I believe both FSs do miss the point. NFS4 especially
so. The question is really not about the most efficient implementation
of POSIX filesystem semantics over the network, but rather whether
POSIX expectations are reasonable in the first place. Plan 9 was
bold enough to simply discount some of those expectations and
the resulting system proved to be much better than what a typical
UNIX provides these days.

Now, what would be really interesting is to see is how the kinds
of requirements that pNFS has can be satisfied with Plan9/9P
approach.

Thanks,
Roman.