From: Kristoffer Haugsbakk <kristofferhaugsbakk@fastmail.com>
Date: Thu, 18 Sep 2025 20:58:22 GMT
Subject: Re: [PATCH v2] docs: update description for '--mirror' option for 'git push'
Message-ID: <78b4ca8a-0440-46a6-8281-a369cb73c304@app.fastmail.com>
In-Reply-To: <97770221-63b8-4375-9787-991717594938@app.fastmail.com>

On Tue, Sep 16, 2025, at 22:04, Kristoffer Haugsbakk wrote:
>>[snip]
>> just before your own sign-off.  See an example like
>>
>> https://lore.kernel.org/git/2f78ab2e28c64f05d807cb9460fa62eb1011a57f.1757879060.git.code@khaugsbakk.name/
>>
>> (by the way, oh, gee, that's an unwieldy long message ID).
>
> It is.  Considering I only send out (git send-)emails at most
> every 15 seconds. ;–(

And I don’t understand why there’s a full SHA1 plus the Unix time.[1]
The Unix time ends up being the same on all patches.  So you need
something more.  But the abbreviated SHA1 for each commit guarantees
that each Message-ID (msgid) will be unique, no?  Why the whole thing?

With an abbreviated SHA1 + Unix time you get:

1. Unique substrings within series from the SHA1s
2. Unique substring among other series from the Unix time unless you
   generate and send out emails very fast

It wouldn’t surprise me if there are people who have taken advantage of
the format to harvest the SHA1 from the message id, even though it isn’t
documented (seems like).  But I think I’ll be fine with an abbreviated
hash in my own tree.

† 1: I did some spelunking and it seems like it has always had this
    format.[2] Or at least in the C form.  I don’t know if there was a
    previous shell or Perl form.
† 2: See d1566f7883f (git-format-patch: Make the second and subsequent mails
    replies to the first, 2006-07-14)

-- 
KriHau