News for hackers. Stuff that matters.
Re:The End of Gmane? by jcranmer on Thursday July 28, 2016 @05:13PM (#12183215)
NNTP is a very nice, simple protocol to implement, both server and client-side.

Well, except for the mess that is RFC 822. At least NNTP decided to go 8-bit safe eons ago, so you just have to shove everything in UTF-8 and you can screw RFC 2047 in the face.

Re:The End of Gmane? by dredmorbius on Thursday July 28, 2016 @10:37PM (#12184737)
What specific issues with 822?
Re:The End of Gmane? by jcranmer on Thursday July 28, 2016 @11:01PM (#12184805)
Major issues:

* RFC 2047 is a pain in the butt

* Handling attachments can be difficult (particularly the inability to safely add binary data)

* The display name/address is both painful in terms of syntax.

* CFWS in general (you can tell who's written an email client by their reaction to the letters CFWS)

* Header fields in general have way too lenient handling, so there's no way to, e.g., guarantee non-ASCII handling (RFC 2047 needs to be handled differently in the Subject: header than from the From: header, and the Content-Type: header does something else entirely).

* Mandatory line-length limits, which means you need to be able to insert line breaks in however you encode custom header values

* No way to get attachments without the body or vice versa

* All metadata needs to be shoved into a header (see above about i18n, line length, and CFWS)

Re:The End of Gmane? by dredmorbius on Friday July 29, 2016 @12:51AM (#12185075)
What would you change given a greenfield opportunity?

What would you add? Remove?

Why?

For myself: the basic RF282 concept of sender, subject, recipient (or newsgroup), date, and references, plus body, seem reasonable.

How best to take these to present capabilities (and needs) is another question.