[Monotone-devel] Re: Few remarks

[graydon hoare]

Bruce Stephens wrote:
> Henrik Holmboe <address@hidden> writes:
>
> > With that said, one thing that comes to my mind to uniquely identify
> > a commit is to use seconds since 00:00:00 1970-01-01 UTC, and the
> > user who commited the changeset. It could look something like
> > "address@hidden:1102092772". It would be unique and still be
> > incrementing (hopefully, or else time would be grinding to a halt).
>
> It wouldn't necessarily be incrementing, since different computers
> don't necessarily have the same idea of what the time is.  (It would
> presumably be increasing on each computer.) 

well, I've actually already implemented this sort of "pseudo-external 
identifier" in the sense of selectors. they aren't always unique, but 
when they are they work fine:

$ ./monotone ls certs address@hidden/2004-12-01T09:11:52
monotone: expanded selector 'address@hidden' -> 'a:address@hidden'
monotone: expanded selector '2004-12-01T09:11:52' -> 'd:2004-12-01T09:11:52'
monotone: expanding selection 'address@hidden/2004-12-01T09:11:52'
monotone: expanded to '15fcbe9c9f966cb320a7d4d83e55f7380f979e7d'
...

perhaps we just need some more tab-completion, or to print out selectors 
in more places in the UI's chatter, so people know they're valid. the 
thing is, in a distributed system you need a content-based security 
primitive like SHA1 hashes in place anyways -- since anyone can be lying 
about the thing they associate with a particular external identifier -- 
so I don't feel particularly wrong about using certs-on-hashes to 
support this.

-graydon