| 09:44 | <smaug> | hmm, dom spec has some animation when clicking on an algorithm (to get the popup where it is referred from) Maybe it has been there for quite some time, but it is a bit annoying |
| 10:19 | <Noam Rosenthal> | smaug: hsivonen any thoughts on https://github.com/whatwg/html/pull/9970#discussion_r1489167432? There was a question whether allowing the new (My sense of logic says it would be fine but I'm reading the room's topic). |
| 11:54 | <hsivonen> |
I can't say for perf for real right now. (In any case, we should file the |
| 11:56 | <hsivonen> | I do expect taking name into account to be more complex and to have some gotcha. Not sure if I can claim a perf effect, though. |
| 11:56 | <smaug> | it feels very surprising to support name |
| 11:56 | <smaug> | it isn't then idref but something else |
| 11:57 | <smaug> | what is the need? |
| 11:57 | <Noam Rosenthal> | Just <a name> like when scrolling |
| 11:57 | <Noam Rosenthal> | there is no need except from being consistent with other links with href that take this into consideration |
| 11:59 | <Noam Rosenthal> | I'm also fine with starting with ID only and then see later if we need to follow up. I don't find the consistency argument strong here, <link rel=expect> is a new thing anyway and you don't necessarily need to be consistent with some old way we used to create scroll-to points in the document |
| 12:00 | <smaug> | yeah, I don't think this is really about consistency |
| 12:00 | <smaug> | <a name> and render blocking are quite different concepts |
| 12:01 | <smaug> | so, I'd rather prefer simplicity here |
| 12:05 | <Noam Rosenthal> | zcorpan: thoughts on ^^^ ? |
| 12:13 | <smaug> | Noam Rosenthal: oh, wait, I change my mind. The syntax is really href |
| 12:15 | <Noam Rosenthal> | smaug: yes, the syntax is href. |
| 12:15 | <smaug> | so it is resolved as an url |
| 12:15 | <smaug> | I assume |
| 12:16 | <Noam Rosenthal> | yes. you can also put a whole URL in there and as long as it resolves to a same-document fragment in the end (given base etc) it would try to match |
| 12:16 | <Noam Rosenthal> | it uses the same matching algorithm for the normal scroll-to (try first without percent encoding, then with) |
| 12:17 | <Noam Rosenthal> | OTOH it doesn't respect "top" etc because that's nonsensical |
| 12:17 | <smaug> | other thing is that rel=expect ... blocking=render feels a bit verbose. I need to read some discussion why we ended up with rel=expect |
| 12:17 | <Noam Rosenthal> | there was a loooong discussion about it |
| 12:18 | <Noam Rosenthal> | you mean as opposed to not having rel at all? |
| 12:19 | <Noam Rosenthal> | (I personally don't mind that if you want to render-block on things you have to be a bit verbose. This is an advanced feature in a way) |
| 12:20 | <smaug> | ...looked, and ok, seems fine. verbose but fine |
| 14:19 | <zcorpan> | Noam Rosenthal: smaug: We do have an opportunity to move away from rel/href for same-document references. I know we had a long discussion already. But <link expect=someid blocking=render> is less verbose and could have a real idref like <label for> |
| 14:22 | <Noam Rosenthal> | it feels weird to have a link without a rel. |
| 14:23 | <zcorpan> | Noam Rosenthal: same for <meta charset=utf-8>, but we did that anyway |
| 14:24 | <Noam Rosenthal> | It does feel that adding all these URL-ish things to this feature doesn't add value, but at this point I don't think those options are materially different. |
| 14:26 | <Noam Rosenthal> | ... as in, their differences relate to options that people are not going to use in practice |
| 14:29 | <smaug> | <link> is already weird in many ways, since what people usually think as link is <a href> |
| 14:31 | <zcorpan> | Noam Rosenthal: I think the base handling is a disadvantage with href - people will just use a hash, and if they put a base element in then the render blocking thing stops working (while label for continues to work). I don't know how common base is, but it seems avoidable. |
| 14:32 | <Noam Rosenthal> | Can you do this? <link expect=id blocking=render rel=stylesheet href="style.css"> |
| 14:32 | <smaug> | (link used to be more like a normal link before https://github.com/whatwg/html/pull/6269 ) |
| 14:36 | <Noam Rosenthal> | zcorpan: I agree that the base thing is error-prone |
| 14:41 | <zcorpan> | Can you do this? |
| 14:42 | <Noam Rosenthal> | OK. I wonder what Domenic thinks about these alternatives (in Tokyo business hours :)) |
| 14:50 | <zcorpan> | Did handling of fragment-only links with a base url change at some point? I recall them being same-document references even if there was a base url, but that has changed? |
| 14:54 | <Noam Rosenthal> | zcorpan: it changed during the spec review; |
| 14:55 | <zcorpan> | Noam Rosenthal: I mean for links in general, e.g. a href |
| 14:55 | <Noam Rosenthal> | oh I don't know then |
| 14:57 | <zcorpan> | demo https://software.hixie.ch/utilities/js/live-dom-viewer/saved/12380 - gecko/chromium/webkit apply the base url |