XSF Discussion - 2022-04-06


  1. mdosch

    > poll: If you ask XMPP client devs in xsf@ and jdev@ to contact you about a vulnerability via email or XMPP, do you think most contacts will be via: > 1. XMPP > 2. email > (it's not even close btw, it's a landslide) For stuff like this I prefer email as I can archive it whereas chat is for ephemeral talk.

  2. mdosch

    > emus, yes, but went to spam (Gmail) Funny that Google loves to put non-google mail to spam while most spam is coming from 🥁 gmail (at least on my server).

  3. emus

    ^^ mdosch but did you receive sonething?

  4. mdosch

    I did.

  5. emus

    ok

  6. emus

    but asking again. what can I do to improve. should I not send from Xsf mail comm mail?

  7. mjk

    The letter probably has DKIM signature (or should have) of the sending server that should match the domain in `From`, and mailing lists are known to botch that, which increases spam score. Maybe there's a way to set From to the mailing list address?

  8. Zash

    The mailing list would probably have to fix it.

  9. mjk

    Right, there's probably nothing sender can do

  10. Zash

    Send from somewhere without DKIM 🤷️

  11. mjk

    Won't help, I think. What I think is happening, is the list server puts its own signature (or strips the existing one), but leaves `From` intact

  12. mjk

    But this is pure speculation, I didn't see the actual letter

  13. mjk

    (Didn't subscribe)

  14. Zash

    "I make the news, I don't read it" 🙂

  15. mjk

    :) I do read, it's just that I get notified via several xmppchats :D

  16. mjk

    If someone is willing to post here the .eml as they received it, it can be investigated further

  17. mjk

    At least the headers

  18. emus

    Well, I am happy to go through the mailman settings if that helps. Peter reviewed and said its fine. But dunno how to not let it look like spam

  19. mdosch

    https://files.mdosch.de/upload/rOWMeaLMpo43Oeo61O8bnJTC/3DjQn0uhQM6vHDVSUG6tDg.jpg

  20. mdosch

    I don't see dkim or SPF. So shouldn't be a problem.

  21. Zash

    Is that all the headers? I'm not sure I really trust that screen.

  22. mdosch

    But maybe k9 doesn't show all headers.

  23. mdosch

    Ah, I read in neomutt and didn't download the full message in k9.

  24. mdosch

    https://files.mdosch.de/upload/GgUqLlucDyZbuCYTazJMAPHd/3cJenQPRQDq5MiBxTW1ivw.jpg

  25. mdosch

    https://files.mdosch.de/upload/NmAJt0oHBsDVWS2rNtacYqVI/MEouVlaNTZiQlCuX57z3UQ.jpg

  26. mdosch

    Skim fail.

  27. mdosch

    Dkim fail.

  28. mjk

    Ah, so the list replaced From and didn't add its own signature

  29. mjk

    Seems it also modified the body ("body hash did not verify")

  30. Zash

    Footer? Often also the Subject is modified on mailing lists.

  31. mjk

    So maybe it'd be worth configuring the list server to add its signature then

  32. emus

    Thanks for disussing guys. I cannot provide a lot of input 😓

  33. mjk

    Basically, the original signature from proton is useless and can be stripped

  34. Zash

    Srsly 90% of all DKIM signatures I see on mailing list posts are invalid.

  35. mjk

    Yup

  36. emus

    Should I extract something from proton?

  37. emus

    Or was it wrong to send it from proton account?

  38. mjk

    But in the case of newsletter it's okay to mangle original sender's data (it's even preferable, I guess), which it already does, it just doesn't sign

  39. mjk

    emus: it doesn't matter where you send from, the list will make the original signature invalid anyway

  40. mjk

    It only should add its own, then verification should stsrt working

  41. mjk

    It only should add its own, then verification should start working

  42. Zash

    Unless you send from somewhere that doesn't use DKIM

  43. emus

    I assume protonmail does?

  44. mjk

    Yes

  45. emus

    https://protonmail.com/blog/dkim-key-management/

  46. emus

    but the issue with this was the same as with Tinyletter. Many people claimed it ends up in spam

  47. mjk

    Zash: well I'm not really sure lack of signatures is any better than invalid ones, in the eyes of spam milters

  48. mjk

    emus: to be clear, valid DKIM signatures aren't a panacea, but imho greatly reduce spaminess of letters

  49. Zash

    DKIM isn't about spam, it's an authentication mechanism.

  50. mjk

    Right

  51. emus

    yes, I understood

  52. emus

    So mailman must ensure DKIM propagation from senders to receivers?

  53. emus

    I can ask two more people in that regard

  54. emus

    if that is what we are looking fof

  55. emus

    if that is what we are looking for

  56. mjk

    emus: no, since it modifies the body, it should put its own signature and modify From accordingly

  57. mjk

    Or at least strip all dkim signatures present in the original (personally I'm not sure unsigned mail is any better, but don't have data to prove it)

  58. Zash

    I think the thing is to tell everyone if the newsletter ended up in the spam folder, to mark it as not spam.

  59. mjk

    The most effective measure!

  60. Zash

    IIRC the mandatory thing to do when you create a new (especially self-hosted) email is to get a gmail account and then mail yourself and mark the email as trusted or somesuch.

  61. mjk

    And maybe add the protonmail address in their addr book

  62. Zash

    So to self-host, you must have gmail. Yay!

  63. mjk

    And maybe add the protonmail address in their addr book (if server-side)

  64. mjk

    But what if the mail I send to myself@gmail isn't marked as spam? (True story!)

  65. emus

    So I summarize: how do we get mailman to either: - put its own signature and modify "From" accordingly - or at least strip all DKIM signatures present in the original sender message (no so prefered solution)

  66. mjk nods

  67. emus

    🙏 thanks for evaluating guys!

  68. mjk

    A third option is don't touch the body, but I'm not sure how reasonable that is

  69. mjk

    A third option is don't modify the body, but I'm not sure how reasonable that is

  70. emus

    I got this input on mailmain: https://wiki.list.org/DEV/DMARC

  71. mjk

    emus: how much are letters' bodies are modified and is it essential to the newsletter? Maybe it's just a useless footer? If non-essential, the easiest first step is probably to just not touch the body and see if that preserves validity of the original signature.

  72. mjk

    Then there's also the matter of SPF, but that's easier to fix

  73. mjk

    Might need to create a test mailing list so as not to wait another month :)

  74. emus

    mjk, thanks. I sent the mail and did "nothing" after that. SPF? Yeah, maybe good idea

  75. mjk

    emus: I mean, how does mailman modify the body? Mailing list usually add some footer, breaking the signature

  76. mjk

    emus: I mean, how does mailman modify the body? Mailing lists usually add some footer, breaking the signature

  77. emus

    there seem to be some html attachement: https://mail.jabber.org/pipermail/newsletter/2022/000000.html

  78. mjk

    > SPF? https://en.m.wikipedia.org/wiki/Sender_Policy_Framework It only involves DNS records, thus easier to setup/amend than DKIM signatures

  79. mjk

    > HTML attachment was scrubbed I'm not familiar with mailing list software enough to tell if the scrubbing is done for the web interface or in the actual letters sent from it. But the html is most likely coming from protonmail. One could try sending a plain text email to see if mailman would pass it along unmodified

  80. mjk

    > HTML attachment was scrubbed I'm not familiar with mailing list software enough to tell if the scrubbing is done for the web interface or in the actual letters sent by the software. But the html is most likely coming from protonmail. One could try sending a plain text email to see if mailman would pass it along unmodified

  81. mjk starts to recall why they hate email

  82. Zash

    SPF should already be set up and I doubt it's of any concern here.

  83. mjk

    Well, the From does contain protonmail.com...

  84. Zash

    That doesn't matter

  85. Zash

    it's the FROM that matters to SPF

  86. mjk

    I might be misremembering how it works then

  87. Zash

    if it's even that

  88. Zash

    the SMTP HELO definitely

  89. Zash

    The From matters about as much as the Subject unless I think it's DMARC that says you can't lie there anymore.

  90. Zash

    Nowhere near the guarantees of the XMPP `from` attribute 🙂

  91. mjk

    Which is why I love and prefer it to email

  92. mjk

    Which is one reason I love and prefer it to email

  93. Zash

    I'm sure anyone who has deployed all the stuff required for modern email would agree (assuming they've seen how XMPP s2s works)

  94. mjk

    > I'm sure anyone who has deployed all the stuff ... or even tried to :D

  95. mjk

    For the record, I actually reread some of that wiki article I linked and, yea, SPF has nothing to do with email headers, so not relevant here

  96. mjk

    Although that `dmarc=fail reason="SPF not aligned (strict)"` needs a separate investigatiom

  97. emus

    Can I somehow serve or support this elaboration? do you want to take a look at the mailman setup? (Zash the password should be at something called Atlas^^)

  98. mjk

    I don't think I'll be able to put enough energy into this (not much of expertise + a hatred for the whole email mess)

  99. emus

    mjk: no worries. many thanks

  100. mjk

    Just leaving some pointers if anyone it motivated to fix this

  101. emus

    👍

  102. mjk

    > many thanks No problem so far :D

  103. Daniel

    is PAM a thing?

  104. Daniel

    are there any implementations?

  105. emus

    thats not about the mailman topic anymore right?

  106. Daniel

    it's about https://xmpp.org/extensions/xep-0376.html aka PAM

  107. emus

    ok

  108. goffi

    Daniel: I'm working on an implementation right now.