ANN of new versions of Mercurial

[Paul Boddie]

On 08/01/13 01:58, Matt Mackall wrote:
> On Mon, 2013-01-07 at 13:05 +0100, Paul Boddie wrote:
>> [*] When asked to give Ubuntu releases a test to see if the bootloader
>> works, to provide an extreme example, I often wonder whether Canonical
>> thinks its demographic is a bunch of tinkerers who don't do anything
>> more than try out the latest release, as opposed to people using the
>> software to do actual work. I also wonder whether they have any testing
>> infrastructure to eliminate basic deployment issues at all. There's only
>> a limited potential in asking people to bear a significant cost
>> (disruption and potential loss of work) in order to save a much lower
>> cost to the project.
> This is an perfect example of my point. Canonical has about 500
> employees, but there are something like 20 million Ubuntu users (most
> paying $0 each). There are many more than 500 models of laptop out there
> so it's obviously an economic impossibility for Canonical to test even a
> tiny fraction of the hardware configurations their users use. So they
> absolutely need people like Tess to help them.

I'm not disputing this. That said, a lot of bugs that can be caught by 
opportunistic testing are generic enough that just running a few test 
machines would pick them up. My bootloader bug anecdote involved a 
pretty normal configuration, and any organisation considering more than 
a couple of use-cases should have been testing for it anyway. (I have my 
own suspicions about what kind of narrow audience Ubuntu is targeting 
these days, but that's a different story.)

> Let's imagine a scenario where .1% of machines are incompatible with a
> new bootloader. If they have 1000 Tesses, they might catch it before
> release. 2000 Tesses and they're much more likely. But Earl is pretty
> much irrelevant. If 20 million Norms randomly upgrade once every two
> years, 100k Norms are going to randomly pick release day to upgrade.
> Without Tess, we'll have 100 Norms with machines that won't boot. If
> we've got 100k Earls updating too, now we've got 200 broken machines on
> day one. Not an improvement.

I think I got Norm and Earl somewhat mixed up, but I also think there's 
a continuum involved. Some forms of Earl want the latest release but 
hold back until it is "done" because they are not merely playing around. 
To persuade these forms of Earl to use release candidates is a hard 
sell, even though any investment on Earl's part does provide him with 
some benefit, potentially huge if there's something that will break his 
working environment.

> This is basically the story of Mercurial, but scaled up by a few orders
> of magnitude.

Well, I wish you the very best in getting Earl to make time for trying 
out the release candidates. Since I'm more in the Norm category myself 
with regard to Mercurial (but not with regard to everything), mostly at 
the more conservative end of that part of the continuum, I won't try and 
change anyone's mind about this, but I do wonder whether the tactics are 
the best ones to encourage earlier testing.

Paul