[TOOLS-DEVELOPMENT] Fwd: Re: Fwd: Re: Planned experiment: A new mailing list for last-call discussions

[Robert Sparks <[e-mail]>]

For conversation:

Keith Moore is looking for something more than currently proposed for 
the last-call experiment. He'd like to be able to subscribe to messages 
about a given document (or topic) by clicking on things in the last-call 
announcement message, and not get traffic for anything he didn't opt-in to.

Thoughts?

(Relevant bits from three messages in reverse chronological order pasted 
below)

RjS



-------- Forwarded Message --------

On 9/13/19 9:40 AM, Robert Sparks wrote:

> So this is the "Topics" configuration under mailman.
>
> I could see automating adding topics, where the topic was a draft 
> name, or perhaps drafts matching the group prefix. But I think there 
> needs to be some analysis of how mailman behaves when the set of 
> topics gets large. 
Yes, I had the same question in my mind.
> Mind if I forward this thread so far to the tools development team for 
> get their input?
I don't mind at all.
> I don't think a custom list expander is something we should take on.
Well, I've written one before.   Back when everyone used sendmail, 
people used to use my bulk_mailer tool to improve performance because 
sendmail (at that time, at least) delivered mail to all recipients in an 
envelope in a serial fashion, so every time an SMTP server was slow to 
respond it delayed mail delivery for everyone else.   Bulk_mailer split 
things up so that no envelope contained more than N recipient addresses, 
and addresses at the same mail domain ended up in the same envelope.   
These days I suspect/hope that MTAs are smarter, and that nobody uses 
sendmail anymore, or at least that sendmail bears no resemblance to what 
it used to.

Anyway the list expander is pretty much an ordinary list expander except 
that the recipient list comes from a database query, or a different 
database query than normal.

The real trick might be arranging for every message to be classified so 
that something can tell which Last Call the message pertains to and 
which topic to assign to it.

Keith

-------- Forwarded Message --------

On 9/12/19 5:43 PM, Robert Sparks wrote:

>> Mailman already supports a very similar feature so it might just be a 
>> matter of adding some UI bits. 
>
> I don't think I'm familiar with the mailman feature you are pointing 
> to - can you send a link or more description?

I've never tried using it, but mailman does seem to have some support 
for letting subscribers opt-in or opt-out to specific "topics".

Alternatively, the mechanism could be implemented outside of mailman by 
doing the processing in the list expansion software and having the 
lastcall list use a different list expander.   The list expander would 
do a database lookup to see which recipients had opted-in to a 
particular last-call action, and only forward the message to those 
individuals.

(or have a separate lastcall-optin list with its own expander)

Anyway I'm pretty sure this would be easy to write, and suspect some 
people would want to use it.   But there's no point in writing it if 
there's no chance IETF would use it.

Keith

-------- Forwarded Message --------


What I'd like to see is a way to easily and selectively opt-in to 
particular last call discussions with a single button press.

For example:  Once a week the list sends out an HTML message with all 
currently open Last Calls.   For each of these Last Call actions, there 
is a button in the message that can be used to subscribe or unsubscribe 
to the specific discussion associated with that last call.    That 
button launches a web page which has the side-effect of changing the 
user's subscription.    Perhaps there should also be a link to the tools 
page for that action, and to archives of discussion of that action.

Also for any Last Call discussion message sent, there's a link in that 
message to allow unsubscribing from the topic (as opposed to the whole 
list).

Mailman already supports a very similar feature so it might just be a 
matter of adding some UI bits.

Probably willing to contribute code as long as I'm not swamped with 
paying work.

Keith