/dev/rob0 a écrit : > On Monday 27 July 2009 16:57:20 mouss wrote: >> /dev/rob0 a écrit : >>> Unfortunately, I have found that many Web programmers don't bother to >>> read RFC's and find out what characters are allowed in email addresses. >>> Many sites will not accept a "+" in your username. I think the old >>> default qmail delimiter, "-", is a better choice for those just now >>> switching to recipient_delimiter use. Another good one would be ".". >>> >>> To name one, I tried to get automobile insurance with GEICO, a large >>> insurer in the USA. If I had access to my old virtual_alias_maps I >>> could find many more who rejected the "+". >> I've seen many sites that refuse '+', but for now, no site that refuses >> '-'. unfortunately, I am not happy with using '-' because: >> >> - many french names have a '-': "jean-pierre", ... >> - '-' is used in mailing-lists > > Understood on both counts, but is this really a problem? As the > documentation for recipient_delimiter states, the full LHS string is > tried first. Your jean-pie...@example.fr could still use > jean-pierre-...@example.fr. I guess the problem occurs only for > j...@example.fr if he wants to use jean-pie...@example.fr. And there, > you have a political / administrative issue. >
it's not politicial. If I tell users that they can use any use...@domain, then jean would be allowed to use any jea...@domain, including "jean-paul". now if jean-paul is hired, then he would get mail for jean-paul, even if this was used by "jean" before. this is a no go. it is not reasonable to tell "jean" that there is an exception list, because I've seen names that I have never saw before. (actually, I suggest to people like "jean-paul" to use "jpaul" instead. but some people love their first names...). > Seems like there is NO perfect choice here ... and all because of > incompetent Web monkeys. :( > yeah. but it's not only monkey developers. in a company I worked for before, we outsourced some web dev to an external company. part of the code was to handle user subscriptions. The first thing I tested was the email address validation. of course, the code rejected '+' (but it accepted a lot of addresses that were obviously invalid ;-) I reported this to my boss, just to hear him saying "I don't want extensions. I want _real_ addresses". we've had a long discussion, but it brought nothing. (fortunately, he later needed me to "improve deliverability"...).
