Re: [gentoo-project] Council meeting 2021-07-11 - call for agenda items

[Sam Jorna (wraeth)]

On 7/7/21 9:25 pm, Rich Freeman wrote:
> On Wed, Jul 7, 2021 at 2:55 AM Sam Jorna (wraeth) <wraeth@g.o> wrote:
>>
>> On 7/7/21 11:59 am, Rich Freeman wrote:
>>> I'd like the council to vote on one of the following, or some other
>>> approach, preferably to get IRC/email/usernames to match once more:
>>
>>> * On 2021-08-01 the LDAP user database will be queried to compare
>>> their IRC nickname to uid/email addresses.  If these do not match,
>>> then the retirement process for that developer will commence.  Users
>>> who are unable to obtain a matching nick should contact <foo> to have
>>> this addressed with Libera/Infra prior to this date.
>>
>> Would this be an ongoing thing?  This also ties into my previous point -
>> if I temporarily switch nicks, would I be in violation?  What if my
>> connection drops/reconnects and I end up as 'wraeth_' and don't notice,
>> would I be in violation and risk retirement (an extreme example, but
>> hopefully you get my point)?
>
> See above.  Nick collisions/etc have always been a thing and I don't
> think having an underscore at the end really creates confusion,
> especially with autocomplete and most IRC clients recognizing your
> nominal nick in highlights.  The issue is that we have a fair number
> of devs (including ones who were among the first to change networks)
> who chose completely different nicks.

Not sure what you mean by 'see above'.

The issue in that example isn't that my nick would have an underscore,
or whatever my client chose as the next available nick, but that it
would be different from what LDAP lists me as since the goal here is to
regulate what nicks people are allowed to use, potentially with the
threat of retirement if it's different.  What if my client picks
'the_wraeth' - it's (arguably) just as trivial as an underscore, but
it'd mess with autocomplete.

You also raise the point that IRC clients typically highlight primary
nicks by default (not to mention you can add custom highlights).
Perhaps the answer should be that whatever nick one is using, they are
expected to respond to their nominated LDAP nick (ie ensure their client
will highlight for it regardless of current nick)?  You might not get
autocomplete, but provided they're present they should still see it.

I'll concede, though, that if they're using a different nick, you
couldn't know they actually got the message.

>> Each of the above three options also mean available Gentoo uids is
>> dependent on whether a given nick is available on Libera.
>
> That was always the case with Freenode.  It isn't an accident that
> they matched before.  New devs would be assigned a uid that matched
> their nick.   So, the namespace was already constrained.

I don't think this is true - bkohler@g.o is typically known on IRC as
iamben, for example, and searching the current developer list[1] for
"iamben" returns nothing.

It may be that people typically chose their IRC handle as their Gentoo
uid, but I'm not aware of any policy that says they must match.

>> Perhaps, rather than forcing either nicks or uids to change, this last
>> option should be properly codified and enforced?  Should there also be
>> some guideline on how long someone can use a nick before it's considered
>> "their nick"?  Alternatively, can LDAP support multiple nick entries?
>
> IMO this is a pretty bad option.  It basically means that the only way
> to go ping somebody on IRC is to first do a table scan to figure out
> what they're called at that particular moment.

How often do you go to ping someone on IRC and find they're using a
different nick you didn't know about since the last time you talked to
them?  You would really only need to look up a nick if the nicks you
knew about weren't present.

> The next obvious step is adding bot commands to facilitate this, so
> now we have more bot spam in our channels as people are constantly
> asking bots whether somebody is already in the channel (since the
> channel member list is now useless), or to relay messages which means
> everybody gets to read the channel twice.  Or people are first posting
> "foo, are you here?" hoping to catch them with highlighting rules.
> Then they say "willikins: is foo here?"

I agree, using a bot to work around this isn't the best solution.  That
being said, from an end-user perspective it would be easy to '/query
willikins whois <dev>' when autocomplete doesn't complete the nick
you're looking for (but obviously would require some amount of effort to
actually implement).

> The whole point of having things like IRC channels is to facilitate
> communication.  Having a level of indirection obfuscates
> communication.  In more rich applications like github/bugzilla/etc
> there are often search functions for things like authors which help
> alleviate this.  Your typical IRC client doesn't have LDAP integration
> so that we can dynamically map nicknames that change according to
> mood.
>
> If we were using something like Matrix then we could have a Federated
> Gentoo identity that is linked to LDAP and so there is never a
> discrepancy.  I agree that there is an argument for the simplicity of
> IRC.  However, having a level of indirection in nicknames goes against
> that argument for simplicity.  If you want a fancy communications
> system that relies on LDAP lookups then at least automate this vs just
> having everybody constantly scanning a webpage.

I do see the difficulty it can cause with communication, but I don't
think LDAP integration with IRC clients or switching platform entirely
is the right solution either.

>> I think at least the issues of multiple/temporary nicks could be
>> mitigated if you replace 'nick' with 'account' - their Libera account
>> name must match their Gentoo uid in order to be eligible for a cloak.
>> That way the cloak matches their Gentoo uid (since the format is
>> gentoo/developer/$account) making them easier to identify, and you can
>> look up a nick in nickserv to see which account owns it (though you
>> can't look up an account and see what nicks it owns, nor if they are
>> currently online).
>
> As you point out, this lets you use whois to figure out who some
> random nick on Freenode is.  It doesn't let you figure out what nick
> to use to ping somebody about something when it happens to be May the
> 4th, June the 7th, August the 13th, or whatever other days of the year
> various random individuals decide to celebrate various random
> holidays.  When you have hundreds of people in an org you can't just
> expect to know that this guy celebrates Ewok day and goes by Wicket
> when the death star is in the waning crescent phase
I think the Death Star has been waning for a while now... :)

>> More generally, how would this apply to people who don't use IRC and
>> thus would never have their IRC nick in LDAP match their uid?
>
> The same way we handle people who don't answer the question in their
> dev application about what their IRC nick is: presumably we don't give
> them dev accounts already.

Is a developer actually required to be on IRC?  My understanding was
that it was technically optional, same as membership on Github or
subscription to most lists.

> It isn't just hundreds of cases of coincidence that everybody used to
> have matching Freenode nicks/Gentoo uids.  This was very much designed
> into the process before.
>
> I realize this whole situation seems a bit silly, but the whole point

I do get the concern in terms of ease of communication, and I'm not
trying to make light of it, but if the intent is to regulate IRC
presence I want to understand it properly.

> of sticking with IRC was to keep things simple, and people changing
> their nicks is actually hindering communications.  Some examples:
>
> 1. I had a conversation with a random person who ended up being a
> Council member, where I wasted everybody's time adding context that I
> didn't realize was unnecessary.
> 2. I've seen devs say on IRC that they need to go talk to somebody
> when they were already on the channel (just under an obfuscated name),
> resulting in delay/etc.
> 3. Most recently I was reading an IRC chat log posted in a separate
> thread where it wasn't apparent until the second reading that somebody
> else in the discussion was being quoted.  I ended up using /whois just
> to confirm this.  Of course, if I were to do this using the list
> archives six months from now who knows what /whois would say.

These are each valid points, but I suspect they're each a symptom/result
of the move to Libera and changes related to that, such as a desired
nick becoming free.  Again, how often is this actually a problem?

> Yes, these are all relatively minor issues, but again, the whole point
> of IRC is communication, and designing our system to practically
> guarantee a future of minor communication problems seems suboptimal.

[1] https://www.gentoo.org/inside-gentoo/developers/

-- 
Sam Jorna (wraeth)
GnuPG ID: 0xD6180C26