Re: HTML export uses anchor ids which change on every export

[Nicolas Goaziou <mail@nicolasgoaziou.fr>]

Hello,

Tim Cross <theophilusx@gmail.com> writes:

> Perhaps I misunderstood. My reading was that none of the proposed
> approaches were complete enough (in the sense they either introduced
> other issues or, while addressing some corner cases, made it much harder
> to address others, broke or failed to cater for other workflows).
>
> I was left with the general impression that solving this issue required
> a significant amount of re-development and a far more sophisticated
> approach for tracking, caching/memoizing IDs and attempting to address
> the issues just by patching the existing code was only going to make
> small improvements while complicating the existing code and making it
> harder to maintain. In short, a significant re-design and
> re-implementation effort rather than application of patches on the
> existing code base is required and until someone can do this work, the
> best approach was to use publish instead of export if link stability was
> required.

I agree on some points, but my analysis is slightly different. In
particular, it seems to me the whole topic is conflating problems. And
the mistake is to look for a single solution that solves them all.

First, _external_ link stability is a solved problem. Users need to use
CUSTOM_ID, no matter what they think about it. I do believe there is no
other automatic way to solve this. Only approximations of a solution,
which will bite you in one way or the other, as you noted.

Secondly, _internal_ link stability is not that important. By
definition, if you're not going to see them, you don't care about what
they look like, as long as they correctly link the expected parts of the
document. Current implementation of internal references guarantees all
internal links do work, with export or publish, but does not go further.
I don't think we need another solution for internal links since they do
the job.

This is not to say there is no problem to solve, of course. Currently,
internal links sometimes leak outside, which understandably bothers
users. Even though there is no ultimate solution for this besides
manually writing every link going to the outside, it may be possible to
mitigate the issue, if users accept to get bitten from time to time.

With that in mind, I think Timothy's solution goes in the right
direction, but, IMO, attempts to solve the problem at the wrong level,
i.e., by trying to unify all links (internal and external) into a single
banner. I'm not convinced this is doable, because expectations are so
different.

However, this kind of solution could be implemented in Org and used by
export back-ends generating external links. For example, Org Export
could provide, e.g., `org-export-punycode', and Org Export HTML could
use instead of internal `org-export-get-reference'. As I wrote already,
Org Export Texinfo does something similar for the nodes it generates.
Since those are meant to be external references, the back-end tries hard
to generate something meaningful (in `org-texinfo--get-node') and, as
a last resort, `org-export-get-reference'. Even though I mentioned it in
the other thread, it didn't attract much interest.

This is, I think, a practical way to improve the actual problem, i.e.,
how to to generate automatically pseudo-stable external links (note: I'm
writing this without contempt).

Regards,
-- 
Nicolas Goaziou