search

xproc / 3.0-steps

Repository for change requests to the standard step library and for official extension steps
10 stars 7 forks source link

Inconsistent format attributes #293

Closed ndw closed 4 years ago

ndw commented 4 years ago

On p:archive, format defaults to zip. On p:archive-manifest, it's optional and the processor is expected to work it out.

I think the format attribute should be treated consistently on these two steps. I don't feel strongly about which approach to take, but I'm guessing that the p:archive-manifest way is a little less typing for users.

gimsieke commented 4 years ago

I think this was a deliberate decision. With p:archive you typically create a fresh archive, and you need to specify its format. If you omit it, you get a Zip. With p:archive-manifest, you extract a manifest from an existing archive. From a pipeline author’s perspective, it is preferable that the implementation infer the archive format for p:archive-manifest, rather than defaulting to zip when no explicit format option is given. So the perceived inconsistency is justified IMO (or there is no inconsistency because the premises aren’t the same).

ndw commented 4 years ago

Okay, but as written, if you do provide an archive and it isn't a zip archive, you'll get an error if you omit the format option even if the archive format could have been inferred from what was provided.

I'd be happier to make the steps consistent, allow the processor to infer the archive format in both cases, and simply say that if no archive is provided and no format is specified, the format "zip" is implied.

gimsieke commented 4 years ago

Okay, but as written, if you do provide an archive and it isn't a zip archive, you'll get an error if you omit the format option even if the archive format could have been inferred from what was provided.

Is it? “If no format option is specified or if its value is the empty sequence, the archive's format will be determined by the step, using the content-type document-property of the document on the source port and/or by inspecting its contents.”

Ah, you are referring to p:archive.

I'd be happier to make the steps consistent, allow the processor to infer the archive format in both cases, and simply say that if no archive is provided and no format is specified, the format "zip" is implied.

Yes, that’s the most user-friendly way. Let’s spec it that way.

ndw commented 4 years ago

Fine