Future of Mercurial?

Augie Fackler raf at durin42.com
Wed Mar 13 15:57:24 UTC 2019


Almost everything you've written here about our language choices is wrong. I'm not sure where you got your information but:

1) Mercurial 5.0 will be a beta release supporting Python 3 (probably with some issues left to fix on Windows - we need help). See https://www.mercurial-scm.org/wiki/Python3
2) We are not "rewriting Mercurial in Rust", at least not on any kind of timescale where predictions are meaningful. We're trying to use Rust instead of C because we're tired of CVEs.

I'm not worried about the future of Mercurial, and it even looks like Python 3 will provide us some nice wins on things like memory use.

> On Mar 11, 2019, at 23:37, Somchai Smythe <buraphalinuxserver at gmail.com> wrote:
> 
> Well, I'll tell you my perspective, but I admit my use case is
> probably not common and my opinions are probably controversial.  Maybe
> it'll spark a response that states the plan going forward, and we'd
> all like know if there is such a plan, and if it exists what it is and
> what the timeframe is.
> 
> Mercurial doesn't really work with python3, and python2 is pretty much
> end-of-life.  Switching to git is the only viable option unless the
> mercurial people fix this, and they seem to be unable or unwilling to
> embrace python3.  This is why I switched all my projects to git, even
> though I much prefer mercurial.  I need something that works
> cross-platform and will still be working in the years to come, and
> anything which requires python2 is doomed.  I even had to throw out
> all my teaching materials and rewrite them for git (that was a
> non-trivial exerciese).  Trying to teach git to students new to Linux,
> well, it's turned out to be about 10 times harder than teaching
> mercurial to them.  But I don't want to teach dead-end technology
> since it won't be useful to them after graduation.
> 
> I'm depressed by the fact that even after all this time the mercurial
> maintainers didn't either learn python3 or rewrite it in straight C,
> but they've hinted they'll go with rust.  My non-teaching systems
> where I once used mercurial are all offline systems and I use
> thumbdrives to move data on and off of them. The rust toolchain
> requires a live internet connection to even build, so I cannot work
> with that toolchain.  How can people even trust something that makes
> it practically impossible to see the source they are actually building
> with without using tcpdump/wireshark to capture the streams?  The
> proponents claim rust is more secure, but who can actually be sure
> since it downloads code you cannot reivew during the build?  It could
> put _anything_ in there and you'd never know it until too late.  And
> yeah, I had to dump firefox for the same reason.  If they had chosen
> 'go', C or C++, they don't have these rust issues and I'd be willing
> to try it.
> 
> Meanwhile, git builds fine without an internet connection, it handles
> the linux kernel fine which certainly builds my confidence it can
> handle anything I'll ever need to do with it, even microsoft has
> switched to git, and it sure seems to me it'll be the last vcs
> standing when the dust settles.  The fact that git is also the most
> difficult and tedious to use is unfortunate, but a price most seem
> willing to pay to get one ubiquitous vcs that builds and runs
> anywhere, builds easily, and is designed for high performance.
> 
> If, and this is rather unlikely, the mercurial team ditched rust and
> embraced python3, which also runs everywhere I care about, I might
> consider switching back, but switching vcs keeping history, tags, etc.
> is so much trouble I suspect most large projects wouldn't even
> consider it once they've completed their switch to git.
> 
> If mercurial wants to remain alive, the maintainers need to deliver a
> drop-in replacement, even if it is a rust-based thing that wouldn't
> work for me, that doesn't use the doomed python2, and it'll need to be
> backwards compatible with the older hg repos.  It could still happen,
> but is it wise to rely on that?  Only you can evaluate the risk and
> make that decision.
> 
> Since the linux distribution I use doesn't even have python2 any more,
> the decision has already been made for me.
> 
> 
> On 3/10/19, Harley Leyton <voldermort at hotmail.com> wrote:
>> -- The following is written in good faith for frank, honest discussion --
>> 
>> I began using hg many years ago, back when git had a horrible UI and didn't
>> work on Windows. Since then, git has become fully supported on Windows and
>> the UI has much improved. hg still has the edge for user-friendliness and
>> cross-platform support, but git has almost 100% of the mindshare and market.
>> 
>> I've been stubbornly sticking with hg for hobby projects, but I almost never
>> encounter anything other than git in the open source and commercial worlds.
>> (I'm aware that hg is used in both, but this is a rare exception.) hg seems
>> to be going very much in the direction of bzr, although we're clearly not
>> there yet.
>> 
>> I'm interested in more positive - but realistic - perspectives.
>> _______________________________________________
>> Mercurial mailing list
>> Mercurial at mercurial-scm.org
>> https://www.mercurial-scm.org/mailman/listinfo/mercurial
>> 
> _______________________________________________
> Mercurial mailing list
> Mercurial at mercurial-scm.org
> https://www.mercurial-scm.org/mailman/listinfo/mercurial




More information about the Mercurial mailing list