3 ms·
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 sa
by jcranmer 10y ago
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.
- dredmorbius 10y agoWhat specific issues with 822?
- jcranmer 10y agoMajor 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)
- dredmorbius 10y agoWhat 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.