Meeting minutes
<skef> present!
Vlad: first topic is Garret's rationale for format changes
Garret: When thinking about incorporating query, it seems we will have to make a change.
Garret: for range request the request has to be a 'get' (which is problematic for patch subset)
Garret: split into two types of server so we don't need a negotiation of protocol. the client/server can be configured differently for each type based on keyword
Garret: a little more to set up but, but more performant overall going forward
Vlad: I think it's a good idea. Some reasons have been discussed, and the underlying problems still exist. It makes sense for the author to determine what kind of server/transfer technology they want to use
Garret: we can still have an 'easy mode' with a third keyword ('figure out the best way for me' mode)
Skef: there's a version where the author may not know what the server supports; there's another where the author needs to decide which is best
<Vlad> There is an open issue on the subject of mandatory support for both methods: https://
Garret: if the author knows, they can set it explicitly (and the server will follow that if it can); otherwise they could use 'auto' and the server would have to figure it out
Garret: the intelligence would have to be in the server to determine which is best
Skef: one option is to have the 'auto' keyword, the other is only having 2 (range request OR patch subset)
Garret: there are scenarios where the server may only support range request
Garret: I was really hoping Myles would be here as this is a significant change in the spec
RESOLUTION: WebFonts WG to ask CSSWG to update tech-incremental definitions in CSS Fonts Module
Vlad: I propose we have a formal request that CSS Working Group from Web Fonts Working Group to update the Incremental Font Transfer definition to allow for the 3 keywords
Skef: It might be good to have something in the spec about what happens if there is incompatibility between what is specified and what the server supports, and in subsequent documents (i.e. load another page that specifies a different transfer)
Garret: one option would be to preserve both states, or possibly just overwrite the first and start using the second
Skef: if we have state, then 'auto' should 'stick' to that method on subsequent requests
Garret: that makes sense to me
Garret: the 'state' should be scoped to the domain, so this potential for mismatch across sites should not happen
Vlad: was hoping we could discuss remaining open issues
Garret: most relate to this topic, so they should be able to be resolved once this topic is merged
Vlad: I will reach out to Myles again and remind him about our discussion back in June (re his participation or appointing a proxy), and find out if he will be attending TPAC
Vlad: that brings us to TPAC agenda
Vlad: we discussed having 2 sessions: 10am-noon, and 2pm-4pm (PDT) on September 13th
Vlad: agenda: open issues with the label 'range request'. We can have an open PR with what we discussed today, and once agreed can merge and update the spec
Garret: We should talk about incorporating 'query' in the spec (in an optional way so it doesn't block the spec)
Garret: I'll try and track down more info on query's status before TPAC
Vlad: any other topics?
Skef: I'd like to talk about font formats (or changes to existing formats) based on the adoption of incremental font transfer. In particular about SFONT tables and the compression of those tables
Vlad: I was thinking about what's next, and was thinking we should think about performance testing
Garret: server performance testing is in a good place, but we definitely need to think more about client side performance
Vlad: we should add that to our discussion at TPAC
Vlad: maybe we can reuse the test harness we used for WOFF and WOFF2
Garret: planning on talking with Dominik about building a test harness in Chrome
Vlad: if he is there, we can invite him to join one of our sessions
John: tangential to what we're doing, but I'm planning to attend the Math WG to see where they are on math notation and support
Vlad: no meetings until TPAC
quit