Re: [xml2rfc] Insufficiency of txt format

[Carsten Bormann <[e-mail]>]

On 2021-02-01, at 17:34, Paul Kyzivat <[e-mail]> wrote:
> 
> I'll be dead before v2 xml is.

I hope not, but this is an indication of the timelines we have to think in here.

Also, our documents have drawn-out timelines, during which the authoring-format-du-jour will change (as will even the “canonical” format).

So txt-level comparison needs to be part of any sustainable long-term strategy.

The interesting part here is normalization.  Rfcdiff has been normalizing the txt all along by removing footers and headers.  It could add normalization steps such as turning box-drawing characters into legacy ASCII.  If thinking about the comparison story is part of the strategy, such a presentation change could become invisible to long-timespan diffs.

Doing XML-level comparison could be aided by converting to a format such as Pyx (line-based ESIS representation) first.  Again, further normalization is crucial (as we have seen in the example of random attribute order already). The most ugly part here is likely to be white-space handling, which is a very sorry story for XML in general, but could be handled in a schema-driven way for RFCXML.  Also, preptool-generated attributes could be removed in the normalization (but they also can carry useful information).

Grüße, Carsten