branch locking mk2.

[John A Meinel]

Robert Collins wrote:
> On Wed, 2006-02-15 at 22:09 -0600, John A Meinel wrote:
>> There will probably be some control files that are still atomically
>> replaced, like revision-history (as long as that exists).
>> Also, I don't know how we are trying to keep knit index files intact.
>> I'm guessing we just expect that adding 1 line will be an atomic
>> action,
>> as long as we add less than X number of bytes.
>> I can think of some things we could do (like adding a checksum +
>> line-length at the beginning of each line, along with a line delimiter
>> before the line (a simple newline could do)).
>> Then we can throw out index lines that don't match their checksum. And
>> the delimiter makes sure that we know where the entries start & stop. 
> 
> Right. The index has to be self delimiting and self verifiable as well
> as not requiring a full download to retrieve revision references. I
> haven't checked if this is the case yet. 
> 
> I.e. in the index we if we have
> ------ 
> record A
> ------
> partial record
> ------
> record B
> parent: revision in record A
> ------
> 
> 
> firstly the partial record must not be able to confuse the parser.
> secondly if I have the revision in record A, I must be able to tell what
> it was without downloading its index entry. (Otherwise we scale
> O(revisions) rather than (O(delta size))

I don't understand what you mean here about needing to 'tell what it was
without downloading its index'.

Do you mean without downloading the knit context, or you mean being able
to skip over downloading that part of the index?

> 
> hg uses the exact sized index records to achieve this AIUI, using file
> size mod record size to determine if there is partial records present. 
> That scales [with a low constant factor] linearly to revisions though,
> which we want to avoid.
> 
> Rob
> 

The problem is my index isn't going to be the same as your index. So I
don't know what I need to skip. My understanding is that we always need
to download the complete index. So we need to keep it small. (hg fixes
it to something like 48-60 bytes).

I don't know how we can prevent scaling indexes to O(revisions). my goal
is just to make the constant factor small enough, that the important
scaling is the O(delta size) for the actual knit content.

John
=:->

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 249 bytes
Desc: OpenPGP digital signature
Url : https://lists.ubuntu.com/archives/bazaar/attachments/20060215/a79ddae9/attachment.pgp