On Sep 19, 2017, at 8:21 AM, C. Scott Ananian
<cananian(a)wikimedia.org> wrote:
On Mon, Sep 18, 2017 at 9:51 PM, Chad
<innocentkiller(a)gmail.com> wrote:
I see zero reason for us to go through all the formalities, unless we want
to really. I have yet to see anyone (on list, or on IRC anywhere at all
today) where anyone suggested (2) was a good idea at all. It's a
horrifically bad idea.
Technically, I did outline the arguments for (2), earlier on this thread.
It was a bit allegorical, though:
https://www.mediawiki.org/wiki/Wikimedia_Developer_Summit/2018/Writing_Tips…
This should be seriously considered, not just dismissed.
I agree the short timeline seems to push us toward reverting to Zend. But
it is worth having a meaningful discussion about the long-term outlook.
Which VM is likely to be better supported in 15 years' time? Which VM
would we rather adopt and maintain indefinitely ourselves, if needed --
since in a 15 yr timeframe it's entirely possible that (a) Facebook could
abandon Hack/HHVM, or (b) the PHP Zend team could implode. Maintaining
control over our core runtime is important; I think we should at least
discuss long-term contingencies if either goes down. Obviously, our future
was most stable when we (briefly!) had a choice between two strong
runtimes... but that opportunity seems to be vanishing. Practically
speaking, it's not really a choice between "lock-in" and "no lock
in" -- we
have to choose to align our futures with either Zend Technologies Ltd or
Facebook. One of these is *much* better funded than the other. It is
likely that the project with the most funding will continue to have the
better performance.
There are other big users of HHVM -- do we know what other members of the
larger community are doing? We've heard that Phabricator intends to follow
PHP 7. Etsy also shifted to HHVM, do we know what their plans are?
--scott
This is what I want to know as well: who is the community of HHVM users outside of us?
What are their own internal reactions? Keep in mind that businesses like Etsy don't
have to support third party downstream users as we do; "I guess we'll use Hack
now" is a more plausible response than it would be for us.
But if others are indeed like us and are balking at the plan to drop PHP, it would be
worth looking into how we can pool resources to "save" HHVM, or create a runtime
that maintains the performance improvements of HHVM while maintaining PHP support.
Incidentally, how much work has been done on incorporating HHVM's improvements back
into Zend?
--
(
http://cscott.net)
_______________________________________________
Wikitech-l mailing list
Wikitech-l(a)lists.wikimedia.org
https://lists.wikimedia.org/mailman/listinfo/wikitech-l