Re: [Nmh-workers] nmh in near, medium, and far-term

[Paul Vixie]

On 12/9/2011 11:13 PM, Oliver Kiddle wrote:
> Jon Fairbairn wrote:
>> Long ago, before mh I used an email system that stored email in
>> files with unique ids, which suggests a way to do this. Switch
>> to storing messages in (say) date-named accession order folders
>> with unique filenames (derived from the message-id, or simply
>> accession numbers) while the mh folders would contain symlinks
>> to these with numeric names as now. So (to continue the example)
>> sortm would not rename any message files, but just rearrange the
>> symlinks.
> If you're going to do that, you might aswell name the files using an
> SHA hash. It's not like mail files change often. The idea makes me
> wonder if git could be used as a mailstore. With a blob for each message
> and tag-blobs for classifying messages. Could be nice if you can push
> and pull mail between machines.

i actually like the idea of putting useful information into the
filename, like the file size with and without CRLF translation, since it
lets us learn things from readdir() without having to call fstat(). not
that we couldn't add a hash of some kind if we need it for uniqueness,
but i don't think we will.