Re: [Monotone-devel] newbie question - SHA1 vs serials

[K. Richard Pixley]

Richard Levitte - VMS Whacker wrote:

> In message <address@hidden> on Tue, 19 Apr 2005 07:40:06 -0700, "K. Richard Pixley" 
> <address@hidden> said:
>
> rich> Why SHA1 instead of serials?
>
> You came to the answer yourself :-) (globally unique, or at least
> close enough).
>  
Ah, but SHA1 isn't the only way to provide this, even in the absence of 
a central authority.

> I assume you've understood that it's really one repository per
> development machine, right?
One or more, sure.  They might also contain zero, but that's not a very 
interesting problem.  :-).

If my laptop is generally ako.noir.com, then I'll probably name my 
projects foo.ako.noir.com, bar.ako.noir.com, etc.  Wherever it happens 
to wake up and dhcp and/or associate, it'll still have repositories by 
those names.

> So a domain name (as opposed to host name) may not be
> enough, and in some cases, maybe not even a FQDN.
Maybe I wasn't clear enough initially.  I meant: name repositories.  Use 
a procedure for naming them which is based on domain names since domain 
names already provide an heirarchically distributed means of allocating 
netwide-unique information.  The central authority isn't responsible for 
assigning unique numbers so much as it's responsible for delegating the 
authority to make such assignments in a unique way.

In the simplest case, the repository may simply be named after the fqdn 
of the machine which hosts it.  I'm not sure whether it should default 
to this case - that's a user interface question and right now I'm 
thinking about correctness not UI.

> rich> as well as providing that oh-so-sought-after monotonically
> rich> increasing number people so love to see
>
> Quite honestly, I don't believe the love for those monotonically
> increasing numbers is that strong.  I thought so too when I started
> using monotone, but I find that the unique features in monotone
> strongly overweigh the increasingly small pain of dealing with SHA-1
> checksums.
>  
Ah, but if we can provide BOTH, wouldn't that surely be better?

--rich