USEFOR at IETF 42 Meeting Notes About 20 people attended, including both Area Directors (Apps). 1. Message IDs: After mild discussion, consensus says this is ready for publication as an Informational RFC. 2. Posted-And-Mailed Discussed in two parts. First, debate over the actual texts: Why do we ignore followups-to in mail messages? We should add a requirement for handling Newsgroups header in mail if Posted- To header is not present. Why is Newsgroup header included in mail reply if it's not posted to newsgroup? Apparently, there is ambiguity in how this has been handled. Some MUA/reader agents have interpreted this as "This is a response to a news article" and others have thought it to mean "This is also being sent via news. ACTION-ITEM: Chris Newman will mail Jamie some feedback on these issues. Part Two: Many issues regarding mixing mail and news The obvious: we need to be careful mixing them. Some folks (Eric Fair, among others) are really opposed to ever mixing them. In general USEFOR seems to agree that end users do want to mix the two. Do we want a header that signfies it's both a mail-n-news header? Problems exist: reply-to-via-email msg from a newsgroup - is this a good thing? A new message sent to both a newsgroup and an email address - is this good? Religious discussion about mixing mail and news ensued. Should use a new header or deal with existing ones and their [broken, noninteroperable] meanings? RESOLUTION: more dicussion is necessary on the mailing list. 3. Cancel-Locks Not much contention here. It appears to be a reasonable draft. People don't want to implement anything much more complex. Developers present agree that writing crypto-code is painful. There were some concerns about the clue-string decreasing security: however, it sounds like these are all ok with decent implementations. Using numbers to identify the algorithm is a Bad Thing. RESOLUTION: the algorithm-number should be replaced by a string. SUGGESTION: the draft should include examples of generating cancel keys. SUGGESTION: it would be good to add some weasel words to allow local sites to cancel articles regardless of cancel locks - we all know that once a site gets an article, it can do anything it leases to it locally, but we should state this explicitly. 4. Main Draft - Article Format Consensus is to move authentication into separate, later draft. There are questions about how much USEFOR should even deal with security issues. RESOLUTION: move all security, authentication, encryption stuff into a separate, later draft. There appear to be IETF problems with S/MIME in drafts. These may be due to time delays waiting for SMIME to pass IESG. Mail-News and other Gateway issues need much more work. ACTION-ITEM: Brian Hernacki will get some text submitted to the WG on this. The Mail-Copies-To header provoked some discussion. Essentially, the consensus is that people who want to get mailed copies of responses should ask for them in [English] as opposed to setting a flag. Copying text should be done by hand/copy buffers to discourage the quote+"Me Too!" syndrome. RESOLUTION: it's a UI problem and we should drop it from the standard. Then we got into the discussion of header classes. Persistant headers - People like the concept of persistant headers, but not the idea of a namespace. RESOLUTION: it's not a class of headers but a characteristic of specific headers. ACTION-ITEM: Dan Ritter will add "persistent" to definitions section. Variant headers - These are items like Path which are modified by transport (and would not be signed). RESOLUTION: postpone this idea until we find an actual need for it. Comment headers - Some folks are opposed to definition because the current practice is different. Also, it makes transport much harder. RESOLUTION: Drop the comment class. All such headers will use Xref-style conventions or else define themselves carefully. RESOLUTION: Drop requirements for relay-agents to deal with Xref headers. There's no need. ACTION-ITEM: Dan Ritter will add definition of comment header to definition section. Experimental headers There appears to be no real gain in using "X-". - since you have the same odds of namespace collisions - except automatic detection - except human recognition. It's a bad idea because - transition to real headers is harder. RESOLUTION: drop it but add a BCP note about passing it in gateways NOTE: Keith Moore will probably not like this, but he left early so... In summary: header classes are a theoretical idea, rather than being useful for implementation. They have all been banished. Issue: date placeholders [DRAFT-IMPLEMENTATION-DATE, etc.] exist in the current draft. Essentially, no one likes flag days. RESOLUTION: talk to whoever wrote this and find another way to deal with it. Interoperability takes precedence over convenience. Finally, we discussed named articles. There were no supporters for named articles present. Trying to deal with it will slow WG down further. It is in the scope of the WG, but not the main document. RESOLUTION: Brad should write a separate draft for this and submit to the WG.