[EMAIL PROTECTED] wrote: > I propose the following for Mozilla user agent strings (I'll give my > reasons below): > Mozilla 1.0 should be released with the user-agent string Mozilla/5.1 > and subsequent releases should be names similarly: > Mozilla/5.11 for release 1.1, 5.12 for 1.2, Mozilla/5.2 for moz 2.0, etc Makes sense. Then it's easy to see what version of mozilla is being used for purposes of HTTP logging. As long as the bit after the decimal point has an obvious connection to the current mozilla version (e.g. 1.2 = Mozilla/5.12 3.4 = Mozilla/5.34) > > Mozilla/6.0 could come at a later date if there's a big enough change to > warrant moving up a major version number. What sort of 'big change' do you mean? Support for a new protocol, support for new markup/scripting languages (WML, SVG, MathML, etc) > > The only proposal I've seen on user agent strings was back in the time > when they expected the Mozilla release to be called 5.0, The version > number that's currently in brackets (X11; U; Linux xxxx; en-us; 0.8.1) > was meant to be the pre release field and only be in pre release > versions. Mozilla/5.x was meant to represent the versions. You mean here: http://www.mozilla.org/build/revised-user-agent-strings.html ? > > Here's my reasons: > 1. Currently Netscape 4.x and below actively invcrement the version > field of the user agent with each new release Mozilla/4.77 is Netscape > 4.77. Therefore it's unlikely to break any browser that's sniffing for > Mozilla/5 Netscape is the only browser that does this. IE doesn't do anything with the Mozilla/ version number since version 4.x. Since IE 4 IE has used Mozilla/4.0 as its user agent. As IE has more in common with Mozilla than Netscape 4.x then it should use Mozilla/5.0. > 2. Major changes were checked in for XUL syntax and the XUL mime type. > There needs to be a way to distinguish browsers released at version 0.6 > (e.g. Netscape 6) with newer browsers. As we don't want people doing > browser sniffing on the vendor specific part of the user-agent string > then incrementing the Mozilla/5.version makes the most sense. Content negotiation is not a valid reason for designing user-agent strings, although that doesn't stop anyone :) > > Remember Mozilla and Netscape 6 have a translate function in the view > menu. This serves some remote XUL to Mozilla. This sends to old mime > type and XUL syntax so that it works in Netscape 6. I'm sure the vendors > want to ensure Netscape 6 still works while supporting Mozilla 1.0 and > NS 6.5, therefore searching for Mozilla/5.1 user agent will indicate > that the new mime type and syntax should be sent. I don't know about XUL in Mozilla, but if changes in the latest version mean the translate function won't work properly in the new releases then something needs to be done to enable it to work while retaining compatibility with Netscape 6.0. Where is XUL used in the translate function all I get is a web page? If possible we should not use User agent sniffing for this though. > > 3. It encourages people who sniff for browsers (whether or not they > should or not, we can't stop them), to use the Mozilla/bit of the UA > string rather than the vendor supplied Netscape6/ bit (assuming that > Netscape 6 is going to be the most popular distribution of Mozilla). If > Mozilla/ increments with each version then people will use that for > version sniffing, if it stays the same then people are likely to sniff > on the Netscape/ paart of the string which will break Mozilla and other > non-netscape releases. Basically you're saying "we can't stop people doing browser sniffing so we should at least encourge them to base their sniffing on Mozilla version rather than the Netscape version"? If so I'd have to agree. If the browser sniffers decide that they're gonna base their sniffers on the Netscape6/version part of the useragent then other mozilla distributors may decided to spoof Netscape 6 making the vendor substring pointless. > > 4. We've been on Mozilla/5.0 for too long now, I've been using browsers > with /5.0 as their user agent since 1998 - mozClassic used /5.0 so > moving to 5.1 makes the clean break away from that codebase (in the > unlikely event someone wanted to suppot mozclassic!!!). > > OK ignore reason 4 :) > bah! Anyway, I've added .netlib to this post as other people have posted user agent comments there but kept followups to .seamonkey
