[css-align] Serialization of the alignment properties.

[emilio]

Should properties serialize the default values? Right now Blink and WebKit do so, which I think it's a bug.

In particular:

align-items: unsafe start;

Serializes in Blink as unsafe start, and only start in FF. I think the computed value should serialize to the shortest form possible.

This is annoying because it causes a lot of WPT failures on Gecko.

[emilio]

cc @javifernandez

[emilio]

There's a similar issue with first baseline, where FF only serializes to baseline.

[emilio]

Also cc @MatsPalmgren, who also had an opinion on this IIRC.

[javifernandez]

Umm, I understand your point about first baseline and I'm working now on matching FF on this regard. I haven't considered the case of tough. Is it the same case ? cc @fantasai to gather more opinions.

I think this is not the only CSSOM related issue we have, and I'm not sure we have a clear agreement about all these issues (see #1041 or #1033 for example) . I wouldn't mind to work on them as soon as we reach an agreement.

[javifernandez]

Some other examples are #418 #878 #1004

[MatsPalmgren]

CSS values should always be serialized to their shortest form. (It's a bug in the CSSOM spec that it no longer explicitly spells that out.) Regarding first baseline specifically, it's explicitly specified in CSS Box Alignment, which says:

Note: Because they are equivalent, and 'baseline' is shorter, the CSSOM
serializes 'first baseline' as 'baseline'.

[fantasai]

What @MatsPalmgren said. Closing in favor of #1564, please re-open if there's some other consideration specific to css-align.

[MatsPalmgren]

It's worth noting though that unsafe isn't the default value in UAs that implements the behavior described under "(no value specified)".
In an UA that implements that behavior, align-items: unsafe start; should in fact serialize to align-items: unsafe start; (IIUC).