Alrighty. Non-starter, unless that can somehow be fixed in the revlog
library (assuming it has a library; or just a format spec?)
Well, in any case, that project is a ways off. Post-WC work for me.
Another option would be sqlite for all metadata and <something> for
all the content. Bit offtopic by now tho :-P
On Thu, Oct 2, 2008 at 4:01 AM, Branko ╚ibej <brane_at_xbc.nu> wrote:
> Branko ╚ibej wrote:
>> Justin Erenkrantz wrote:
>>> On Wed, Oct 1, 2008 at 10:06 PM, Eric Gillespie <epg_at_pretzelnet.org> wrote:
>>>> close to being competitive with DVCS. You need more in your
>>>> list, starting with merge tracking.
>>> As I remember it, DannyB's argument was that if we switch to a revlog
>>> format, merge tracking comes essentially for free. That was the
>>> takeaway I took, but that was like a while ago now... =) -- justin
>> Eh, not quite. Revlog as it stands today stores two ancestors per node,
>> which is fine except for the cherry-picking case, where it breaks
> Oh, also the revlog-of-today (the implementation) has one *big* problem:
> it expects to be able and willing to read whole files in memory. Perish
> the thought.
> -- Brane
Received on 2008-10-02 19:08:46 CEST