Re: bug#29108: 25.3; ERC SASL support

[J.P.]

Amin Bandali <bandali@gnu.org> writes:

>> Right. Too cryptic. I've adjusted things in the second patch but am
>> happy to redo/revise, as always. (The first patch contains a bug fix.)
>>
> [...]
>>
>> Good call. I've attempted something like that in a separate "examples"
>> section (2nd patch). I'm hesitant about the last, "multi-network"
>> example, though. It sort of implies we're committing to supporting
>> let-binding as a means of specifying per-network local-module options,
>> going forward, which maybe also puts us on the hook for (eventually)
>> providing a mechanism to make options bookkeeping easier for would-be
>> local-module authors. OTOH, neither of those is as yet a realistic
>> problem.
>>
>> Speaking of maintenance burdens, I think `erc-sasl-password' is too
>> overloaded and unwieldy, particularly WRT the "non-nil symbol" form. And
>> falling back on `:id' is redundant because `erc-auth-source-search'
>> already does that. So, as penance for my ugly API design, I've attached
>> a (third) patch that tries to corral some of the crazy by adding an
>> optional auth-source query function to house the more nuanced
>> functionality (for those actually wanting it) while sparing everyone
>> else the needless complexity. (That's the idea, anyway.)
>
> Thanks!  Yeah, I don't really see supporting let-binds as too big of a
> potential future burden.  Worst case scenario, if/when local modules
> are fully implemented and we and/or module authors find supporting
> let-binding impractical, or there are clear advantages in not having
> them, we could drop them in a major version bump (ERC 6.0 anyone?)
> along with any other potential breaking change we might want to make.

Agreed (let's break it all)!

Seriously, though, I really appreciate your taking the time to review
these.

> Also your simplification sounds good.  I pushed the v2 of all three
> patches from your other reply to emacs-29.

Sweet, thanks so much.