06:35
<jyasskin>
annevk: Sorry about the @-references in commits. Is it better to mention your name without the @, or does that still spam you?
06:51
<annevk>
jyasskin: that would be fine
06:52
<jyasskin>
'k; I'll rewrite them.
10:01
<tobie>
TabAtkins: filed https://github.com/tabatkins/bikeshed/issues/938
17:03
<annevk>
tobie: is nobody funding the infrastructure?
17:20
<tobie>
annevk: funding what?
17:59
<annevk>
tobie: computing cost, maintenance
18:00
<tobie>
annevk: of what? Specref, wpt-bot, and pr-preview?
18:00
<annevk>
Yeah
18:01
<tobie>
annevk: no.
18:02
<annevk>
At some point we should get that fixed
18:02
<annevk>
Makes no sense for orgs to pay fees to standards bodies and infrastructure getting none
18:03
<tobie>
annevk: well, the real cost is dev + maintenance.
18:03
<tobie>
annevk: computing costs are super-low/free
18:03
<tobie>
annevk: I think I pay USD 7/month for specref hosting
18:04
<tobie>
MSFT helped with the some of the dev costs of Specref.
18:04
<annevk>
Sure, yeah, both ought to be compensated, and if we need more memory that should be doable too
18:07
<tobie>
so did Google with pr-preview (the chunk that related to making it work for WebIDL, not the generalizations to extend its usage to other specs.)
18:08
<tobie>
annevk: the specific issue with more memory in the specref case, is we're outgrowing the available heroku setup
18:09
<tobie>
annevk: so we can't buy more memory, we need to either make the system horizontally scalable (dev costs)
18:10
<tobie>
annevk: or we need to use a different system (config cost + learning cost)
18:10
<tobie>
annevk: hence the logical move seems is to try and lighten the burden on Specref
18:12
<tobie>
annevk: agree 100% that SDOs should spend a larcger chunk of their fees on infra work
18:13
<tobie>
annevk: it's one of the things that really irritates me.
18:14
<tobie>
annevk: that said, some of the stuff pr-preview does is made possible because of underlying infra provided by W3C (the Respec builder and the html differ)