Hi Nikita,
I am carbon copying this response to [EMAIL PROTECTED], a discussion group for mail-archive.com. This is mainly so I can put all the good comments in on place. Thanks for the feedback! >1. (a small one, really) You mention in the FAQ that your software >sometimes gets confused and archives messages CC'ed to two mailing lists >to the same list twice. Is your list sorting based on the header To >and CC header fields, or on the 'From ' envelope field? The latter is >usually much more reliable, works for mosts lists, and would eliminate >the above problem. Actually, I'm using a relatively complex set of heuristics to sort mail. For the last six months, whenever something went wrong I'd refine the heuristics. About 10 headers are looked at, with different priorities, etc. *BUT* I am not currently using the From envelope field. Stupendous suggestion; can't believe I missed it. Thank you. >2. While mail-archive is currently useful for archives of a list, I >think it has a potential use letting people _read_ mailing lists, >without having to subscribe to them. The important part is keeping >track of which messages are already read. A simple way to do this is to >make an NNTP server which would serve messages from the mail archive, >with groups such as mail-archive.lists.linux-kernel, etc. Then any user >can just set their NNTPSERVER variable appropriately, and read the >mailing lists as newsgroups. I think this should be a not-so-difficult >modification to MHonArc. Hmmm. First, let me mention I don't know the NNTP protocol, so am by no means an expert. However, I do know there are mail->news gateways out there. Also, while MHonArc is supremely flexible, I don't know if it is appropriate for generating NNTP (again I am ignorant of the protocol) You may want to look at the list-admin archives for mail->news info. Looks at first glance like a separate but relevant project. >Alternatively (a more difficult project), some WWW software which hands >out cookies and keeps track of each user's read articles. This doesn't >scale as well, since the state is kept at the server side, as opposed to >distributed to the clients, but a WWW interface has some advantages. I'm assuming this would be used in dynamic indexes? www.mail-archive.com uses static web pages, so browsers already know about viewed messages, and change the colors on links. What do we gain by tracking readers at the server? I'm pretty focused on keeping the service simple, but for really significant gains in user-friendliness would be willing to increase complexity. Jeff
