On Sep 17, 2004, at 5:05 PM, Steve Greenland wrote:
> On Fri, Sep 17, 2004 at 10:12:07AM -0400, Scott Palmer wrote:
>>
>> I'll just put in my 2 cents...
>
> <steve add his two cents to the pot> "Call."
>
>> Often the timestamp of the file is the only meaningful indicator of
>> which "version" of a particular file I have. It is a property that is
>> available to anyone, even end users of our software can easily check
>> dates to see if they have an old file or something like that.
>
> My experience[1] is that if the only indication of file version is a
> timestamp, then you have nothing more than misleading, useless piece of
> noise. Timestamps are *way* to fragile to place any trust in. Oh sure,
> they work often enough to suck you in, and then blow up at the critical
> moment (e.g. confirming over the phone that the client's version of
> the library is newer than the buggy one, and spending the next three
> hours (invariably Friday from 5pm to 8pm :-() trying to find some new
> problem).
>
> I'm vaguely disturbed by the faith some people have with them.
But it is precisely the fault of Subversion an other software to
*preserve* the timestamps that causes this very problem!!!!!
Anyway I usually use the timestamps to show that a file is the same as
another more often than using it to infer that it is different.
> And does MS really base their hotpatch scheme on file dates, instead of
> looking at the version resource or some other signature?
But the timestamp should be just as good if not for *broken* software
that changes the "modification time" when it shouldn't. MS supports, a
creation time, modification time, and last access time. Generally, the
creation time changes when a copy is made, but the modification time
does not.
Of course in this case it is not suitable for the very reason you
mentioned (security).
> [1] If anyone cares: 15+ years of development and support work on a
> variety of machines from DOS to OpenVMS (where each file had, IIRC, *4*
> timestamps, each updated in various known but baroquely complicated
> circumstances.)
I've got you beat by almost 5 years. :)
Scott
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org
Received on Sat Sep 18 03:19:16 2004