Re: [xml2rfc-dev] XML evolution (Re: v3 <seriesInfo> vs IMPNOTES vs canonical XML)

[Julian Reschke <[e-mail]>]

On 19.12.2019 18:26, Carsten Bormann wrote:
> On Dec 19, 2019, at 12:38, Julian Reschke <[e-mail]> wrote:
>>
>> invalid
>
> That points out a flaw of “validation thinking”:

Well, I was referring to the concept of document validity (being valid
with respect to the published grammar and potentially additional prose
constraints).

> There may be structures that are undesirable (deprecated, etc.), but should not trigger a validation failure (i.e., giving up processing).  If we want to evolve RFCXML over time, we’ll need this sooner or later.
>
> I believe that structural immutability of published RFCs may be too high a goal; we of course want to keep a history of what has been published, but with something as complex, error-prone, and evolving (as in moving target) as RFCXML, it also may be necessary to fix bugs.
>
> (I also think we should minimize those cases, and I’ll recognize that the current process is not conducive to minimizing them — that is a separate issue from the one I bring up here; it just makes being able to handle deprecation and fixes even more urgent.)

Evolvability is indeed important. We can experiment with new features
(optimally in an extension namespace), and, for the time of the
experiment, down-convert when publishing canonical XML.

I also have no problem revising the vocabulary more quickly, but right
now the focus should be fixing the mess we're in, not adding more to it...

Best regards, Julian