Bundle2 last call

Gregory Szorc gregory.szorc at gmail.com
Thu Feb 5 13:58:22 CST 2015


On Thu, Feb 5, 2015 at 11:54 AM, Eric Sumner <ericsumner at fb.com> wrote:

>
>
> On 2/5/15, 11:46 AM, "Durham Goode" <durham at fb.com> wrote:
>
> >
> >On 2/5/15 10:07 AM, Matt Mackall wrote:
> >> On Thu, 2015-02-05 at 12:46 +0000, Pierre-Yves David wrote:
> >>> On 02/04/2015 09:59 PM, Matt Mackall wrote:
> >>>> On Wed, 2015-02-04 at 21:57 +0000, Pierre-Yves David wrote:
> >>>>> On 01/20/2015 08:33 PM, Matt Mackall wrote:
> >>>>>> Ok, I've looked over the bundle2 discussion to date and while I see
> >>>>>>some
> >>>>>> things that could be better, I don't see anything that I think will
> >>>>>> cause us significant pain in the future. So I'm inclined to move
> >>>>>>forward
> >>>>>> with turning on bundle2 in its current form at the start of the 3.4
> >>>>>> development cycle. So if there's anything you think MUST be fixed
> >>>>>>before
> >>>>>> we move forward, now's your chance to convince me.
> >>>>> Ultimate call, I'll probably send a lazy patch with s/HG2Y/HG20/g
> >>>>> patches before the end of the week otherwise.
> >>>> We might consider just letting the 2Y stand, so as not to invalidate
> >>>>any
> >>>> of our field testing with churn.
> >>> My english foo fails short here. I'm not sure what you are trying to
> >>>mean.
> >>>
> >>> - If you are afraid that renaming the format break stuff I see this as
> >>> an argument to change it. It should not and I would like to be warned
> >>> early if it does,
> >>>
> >>> - If you are afraid that currently deployed experiment using HG2Y get
> >>> confused, I think it is easy to keep compabilitly between the 2 for a
> >>> couple of month. Especially if the underlying format see no changes.
> >> I think it's just easier to not change. But I don't have a strong
> >> opinion on this one.
> >>
> >+1 on having a grace period where both are supported.  We have
> >non-trivial bundle2 deployments and a hard break would be a pain in the
> >butt.
>
> I'll throw my hat in for not changing it at all, as I don't see a benefit
> in distinguishing between files written before and after it becomes
> officially supported.  If we do change it, I'm +1 on a grace period as
> well.


If nothing changes, can't 'HG2Y' just be forever aliased to 'HG20'? The
grace period would only be needed for advertisement in the wire protocol,
not at-rest data. Or is that already what is being proposed?
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://selenic.com/pipermail/mercurial-devel/attachments/20150205/178a9239/attachment.html>


More information about the Mercurial-devel mailing list