Re: [hypermail] hashed filenames patch commited to CVS

From: Peter C. McCluskey <pcm_at_rahul.net_at_hypermail-project.org>
Date: Tue, 1 Oct 2002 10:28:43 -0700 (PDT)
Message-Id: <20021001172843.8E3E68B94_at_yellow.rahul.net>

 jose.kahan_at_w3.org (Jose Kahan) writes:
>Hello Peter,
>
>Thanks for your feedback.
>
>On Mon, Sep 30, 2002 at 11:10:36AM -0700, Peter C. McCluskey wrote:
>>
>> I'm not very happy with using hope as a means of avoiding problems.
>> Is there a good reason for not detecting and fixing collisions?
>
>Well, not any good one I can think of! Not using a sequential
>naming system means that it's hard to fix collisions. If you do that,
>you go back to the sequential naming system.

 I guess I don't understand what's wrong with a sequential naming system. It seems to me that all you really want to do is divorce the file names from the chronological ordering of the messages. That doesn't seem to require hashes or longer filenames. As long as we keep a mapping between filenames and the numbers used for the mbox or chronological sequence, and don't use the filename sequence for anything other than guaranteeing filename uniqueness (i.e. if we insert a message into the middle of an mbox, we would create a filename with a number higher than the highest filename), then we seem to get what you want without making filenames uglier and without having to worry about how often the hash creates collisions.

-- 
------------------------------------------------------------------------------
Peter McCluskey          | Free Jon Johansen!
http://www.rahul.net/pcm | 

Received on Wed 02 Oct 2002 05:01:28 PM GMT