Greg Wilkins

Strangely enough we are also failing `org.springframework.web.reactive.function.MultipartRouterFunctionIntegrationTests#proxy`, which undertow also is failing. We fail this one with a leak and read after the channel is complete. Is there an issue...

Note that eclipse jetty has had a proprietary extension to `sendError` for some time to support 102 Processing and now 103 Early Hints. We also treat -1 specially as a...

In Jetty we already have a mechanism to order and exclude SCIs (including those on the container classpath). Currently we do it with context attributes that can have an explicit...

> @gregw Do you have a proposal for this scope of work? Not particularly. As I said, jetty uses context attributes, but that is driven more by not creating proprietary...

I'm not a big fan of absolute priorities like that. You can end up with 99.9999 vs 99.99999 as everybody tries to be first or last, especially when baked into...

> This is about SCIs coming from JEE spec. E.g. Faces, CDI, WebSocket. So there's a predefined set. User/3rd-party-library defined ones can safely be ignored because JEE impls usually won't...

Fragment ordering will never be enough to order SCIs as fragments can have multiple SCIs and SCIs provided by the container are never in fragments. So we need a different...

> If this is going to make it too complicated in your opinion then we can also go for the annotation proposal. No problem. Whatever works the best for Servlet...

@markt-asf Are you sure there is verbage in the current spec saying that the container should truncate a too-large write? Consider the following code: ```java String pileOfPoo = "💩💩💩"; String...

@stuartwdouglas I think there is a reasonable interpretation of the spec in all cases: + If uncommitted and a write/print is received with: a) too few bytes, then potentially the...