Version controlled (git) Maildir generated by Dovecot

[Adam]

Hi David, 

Please find answers bellow.

Kind regards,
Adam

---------- Původní e-mail ----------
Od: David Myers <david.myers.24j74 at gmail.com>
Komu: Adam <adam.ranek at seznam.cz>, Dovecot Mailing List <dovecot at dovecot.org>
Datum: 7. 10. 2020 8:44:28
Předmět: Re: Version controlled (git) Maildir generated by Dovecot
Hello Adam, and the dovecot list

> Just a question, I hate to pollute the thread, so feel free to push these
> questions into a new thread if deemed necessary. So as you can guess I'm a bit
> of a newb here, so rather obvious questions are about to arrive....
> 
> As you are using GIT for your archive (which is a cool idea by the way) I'm
> sure you are well aware that not all files types play nicely with version
> control, my question therefore is : How do you plan to handle attachments ?

I use git for everything including for example LibreOffice / Word documents. Git works just fine with binary files. You can't use text tools like "git diff" but... it works.

> Also, although I appreciate the idea of using git, emails generally don't
> 'change', but I guess that also depends on how you are storing them (single
> email with links to previous / next ... etc, or as a single big file for each
> specific thread). Although this is hitting my limits of understanding for how
> dovecot works, so I probably need educating on this (a pointer to the docs would
> be good).

As I mentioned in the first e-mail, I configured Dovecot to use Maildir format -> each e-mail is a single text file. Mail body + attachment(s) are in the same file, attachment(s) are Base64 encoded.

> You seem concerned regarding the files that you are ignoring that you will need
> to 'recreate them', so why not do a complete git add . prior to adding them into
> the git ignore, then you have an initial state for those files too.

But I don't want to store files that can be regenerated. I don't want to backup stuff, that doesn't have information value.

> Final thought, what advantage do you envisage by using git as opposed to simply
> using a filter to select the files over a certain age, and place them into a
> zipped TAR archive ? Although I guess you could eventually zip the git archive
> too, and in the interim it would remain searchable by your users mail clients
> whilt in git.

I like to use git ;-). Tar will work just fine.

In this use case the only real benefit of git is that it never forgets. Unless I delete whole .git directory, I can make a mistake, delete some e-mails (files), commit changes and rollback. I can't rollback if I delete tar archive.

> Thanks in advance, and apologies once again for polluting your question with my
> own.

David