RFC: scaling-inventory rework current thoughts

[Robert Collins]

On Wed, 2008-08-13 at 16:56 +1000, Martin Pool wrote:
> 
> I definitely agree we should not just assume this is sufficient.  In
> particular it doesn't answer how we will store file_id->name mappings.
> 
> I don't find the cases of very large or very skewed trees enormously
> convincing because that shape is going to tend to bite the user during
> other operations such as working on the checkout.

Thanks for the feedback, I agree with what you said; we could add two
conditions - 8 and 9, respectively having a cheap-to-update strong
validator, and can substitute for testaments in an efficient manner; I
think they are separate concerns and probably we need to remove derived
data to achieve 9, but can do 8 sensibly.

As for large or skewed trees, note that ext3 and ntfs and other trees
have internal tree structures within directories specifically to deal
with the large-directory case. (So we could split by directory and
further within directories, but this is ~= to 'not splitting by
directory' IMO).

-Rob
-- 
GPG key available at: <http://www.robertcollins.net/keys.txt>.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: This is a digitally signed message part
Url : https://lists.ubuntu.com/archives/bazaar/attachments/20080813/e48b0590/attachment.pgp