Heho,
> We spent months working out the details, including why it uses HTTPS rather 
> than DANE, on public mailing lists in the IETF. (I would have preferred DANE, 
> but the choice of HTTPS was not made casually.)
To clarify, my comment did not want to pull the found consensus into question, 
and I do not doubt that there are good reasons _for_ HTTPS at this level. This 
comment relates more to issues in the operational implementation I encountered 
when I (recently) implemented MTA-STS; The most pressing one that the MTA-STS 
policy is bound to the recipient domain, which means that I can not simply roll 
it out for all my MX, iff they are accepting mail for domains where I do not 
control the DNS, and my favorit MTA not supporting MTA-STS, because they do not 
want to include an HTTP client. However, I also assume that such issues were 
indeed discussed, and a tradeoff happened.

> If this is something you care about, where were you?
This one hurts, mostly because I know where I was, and also know where I rather 
would have been, doing what. ;-)

> I have certainly run into plenty of people who've had trouble getting their 
> mail into Gmail, loudly announced that GMAIL HAS BROKEN MAIL FOR EVERYONE IN 
> THE WORLD, then I take a look and say "do you know what SPF is?"  "No, why do 
> you ask?"  Sigh.
I think this gets to the core of why centralization for many things is 
succesfull in the first place (leaving the whole good/bad/intention discussion 
out of it). Running systems is not easy; Especially for basic infrastructure 
(which email is), it should just _work_. Then again, over the past three 
decades, it also got _a lot_ more complex (see [0] for my favorit summary on 
that); There is also some work I was involved in which I hope to be able to 
share with the list by the end of the month, goin into the direction of "good 
setups" w.r.t. mail hosters, and the results align pretty much with your 
observation there.

However, it also circles back to the age old question (among people sceptical 
of centralization) of how we can have more distributed infrastructure, without 
having it a) constantly break, b) crappily maintained, and thereby c) causing 
more issues than it solves. At the moment, I sadly do not yet have a good 
answer for that, and I suspect that it won't have a technical answer at all. 

With best regards,
Tobias

[0] 
https://dataswamp.org/~solene/2021-07-09-obsolete-feeling-in-the-crossfire.html

_______________________________________________
mailop mailing list
[email protected]
https://list.mailop.org/listinfo/mailop

Reply via email to