From: spwhitton@spwhitton.name   
      
   Hello,   
      
   Ian Jackson [10/Feb  3:24pm GMT] wrote:   
   > Package: dgit-infrastructure   
   > Version: 14.7   
   >   
   > Empirically, now, this URL   
   >   
   >   https://lists.debian.org/msgid-search/?m=E1vmXg2-00000005atw   
   2ivJ@tag2upload-oracle-01.debian.org&firsthit=1   
   >   
   > goes directly to the relevant tag2upload report.  So the "reported by   
   > email" thing on the web page could link directly to the message, which   
   > would be great.   
   >   
   > This means the Manager would need to know the Message-ID.  There are   
   > three basic ways we could organise this between Manager and   
   > Oracle/Builder.   
   >   
   >  1. Since Oracle/Builder always sends zero or one mail to the user (or   
   >     tagger), it could use a predictable Message-ID for that email.   
   >   
   >     Predictable Message-IDs might have some minor adverse security   
   >     implications, though.  For example, an attacker could pre-send a   
   >     bogus message with that ID to the mailing list and thence to   
   >     USENET gateways.   
   >   
   >  2a. The Manager could generate a Message-ID for the user report, and   
   >     tell it to the Oracle via the o2m protocol.   
   >   
   >  2b. The Manager could generate a Message-ID *prefix* for *all* emails   
   >     from the Oracle/Builder for this attempt, and the Oracle/Builder   
   >     would append a fixed component (the "status" perhaps, or perhaps   
   >     merely append ".starting" for the non-final message).   
   >   
   >     This would mean the final disposition email Message-ID would be   
   >     predictable given the "starting" email; see above.   
   >   
   >  3. The Oracle email-generation code could explicitly generate and set   
   >     a Message-ID for every message and, for the case of the to-user   
   >     final email, report that to the Manager.   
   >   
   > I think I would like to do 2a or 3.   
   >   
   > I expect that 3 will be easiest, but probably the best approach would   
   > be to try to implement it in oracled, as an experiment to see what's   
   > most convenient.   
      
   (2a) seems slightly cleaner for future debugging because then msgid   
   generation always happens on the Manager, instead of possibly happening   
   in two places?   
      
   I guess for (3) we could include the message-id as a new field in the   
   existing 'email reported' line of the protocol.   
      
   --    
   Sean Whitton   
      
   --=-=-Content-Type: application/pgp-signature; name="signature.asc"   
      
   -----BEGIN PGP SIGNATURE-----   
      
   iQJNBAEBCgA3FiEEm5FwB64DDjbk/CSLaVt65L8GYkAFAmmMYnAZHHNwd2hpdHRv   
   bkBzcHdoaXR0b24ubmFtZQAKCRBpW3rkvwZiQMG2EACJtGlCA92stALnO3z+8bqW   
   B4FLujeyMM/F4vlVtI4U1biE78yCt398RQ6hUDMLLz5SPMOS44/Wepx8Kzs910Ud   
   IenjFOHqf2+AChax9WflE2N7H5R4txskJXuSMyF+FtYdmegX4IFEfwX/NHjETHiV   
   i8yf9UuXNl4yDDIsOoBc3TC9aWM9VBU5e6MLFqNPktmkKH2HVIPzKpH5Y1oVSLDQ   
   kQHXSNATAfNEgicJOF7Be6c6gVpebWQ0iG/KqCvAZJt5855OgEQqzjr4X0tIWWNY   
   uCJdSznzAMADRBkr0Vd9dvoPcLGeCw+MZQbobtsn6ERfA3QYYpyBjM2ASvuk9vhu   
   6akSEzT43wGtIJWaFGLqZwvROvI2wvcgu90mAVWyBaDW0/0bKO2O4XtZj1WVsiXj   
   tdTQWdbzfq79G/LnMYktYzTdzmhEa6xkWobALgYql7iYzOtRcrv34VBp8mUC5uQG   
   chX6GkDyvZD429Ykg0ROCdME4GyaQSdeSlrOLZHOwt0nJTx3u8Ph0/xQbiyxwRUN   
   2Sf511B8Vf5Jrandu71Cb8cBOo4fLY4elkTQ7OHrhVi8nFqkY1GqHfPT/F0FlqWN   
   Az+5wAURG21aj+NGEA8e76l7wydIbCVewCYa6ZJ0WhzN+Su8p7gVYCXcW9TvL/GS   
   gHSIZmIwfqUMJHIvzBuGNA==PH2k   
   -----END PGP SIGNATURE-----   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)