[OpenAFS] Windows cache problem revisited...

Phil.Moore@morganstanley.com Phil.Moore@morganstanley.com
Thu, 4 Dec 2003 16:01:25 -0500


>>>>> "Jeffrey" == Jeffrey Altman <jaltman@columbia.edu> writes:

Jeffrey> We are all aware that the design of OpenAFS for Windows is
Jeffrey> not appropriate for the Windows 2000/XP/2003 environments
Jeffrey> with multiple (perhaps simultaneous) users per machine with
Jeffrey> roaming profiles and integrated Kerberos authentication.  The
Jeffrey> work that must be done is substantial and for the most part
Jeffrey> should not be done piece meal.  Instead, a significant
Jeffrey> re-design of the architecture must be performed to better
Jeffrey> integrate AFS into Windows and its local security
Jeffrey> architecture.

No disagreement on that point.

Jeffrey> I certainly understand your frustration.  OpenAFS on Windows
Jeffrey> has certainly been the hand me down sibling of the Unix
Jeffrey> counterpart.  Rants will not change this.  The only thing
Jeffrey> that will is a commitment of resources and money.  A really
Jeffrey> rough estimate of the work to be done falls somewhere between
Jeffrey> $750,000 and $1 million range.  Clearly none of the
Jeffrey> individual organizations which use OpenAFS on Windows are can
Jeffrey> afford to put up that kind of money.

Um...  Wrong.  If someone could show me a design, and a rough project
plan, and demonstrate that the manpower was available to implement it,
I can get you the $$.

As goes AFS, so goes the Morgan Stanley UNIX infrastrucutre.  Now that
we're rolling out the Windows AFS client, so goes the Windows plant,
too (soon).[1]

However, I suspect that the real cost of reimplementing the Windwos
client is more than that, but I'm a cynic.

This doesn't mean that I'm going to cough up a mega-buck so people can
hack.  That level of committment would require some kind of
organization to manage the development process, and own it.

I'm sending this reply because we *ARE* very serious about improving
the Windows client (and in fact, one our own guys has submitted many,
many patches for the Windows client over the last year).  

Let me take this offline, and discuss it with the rest of Council of
Elders, and maybe we can get Sine Nomine, CITI, or someone else to
front the development.

Phil

[1] Ever since seeing Matrix Revolutions (which I hated, BTW), I have
    been having this recurring nightmare about the uncoming deployment
    of 30,000 Windows clients, and the subsequent load they are going
    to place on my UNIX AFS server infrastructure.

    You know that scene where there's total silence in the dock, after
    the EMP wipes out the first wave of attack?  You hear this buzzing
    coming down the tunnel, and then about 100,000 sentinels burts
    through the hole and swarm into the dock area?

    My servers are the human defenses, and Windows AFS clients are the
    sentinels...