Re: Re: Revert session_serializer_name(), session_gc()

On Sat, Mar 15, 2014 at 2:10 PM, Andrey Andreev <narf@devilix.net> wrote: > On Sat, Mar 15, 2014 at 6:36 AM, Yasuo Ohgaki <yohgaki@ohgaki.net> wrote: > > Hi all, > > > > On Sat, Mar 15, 2014 at 1:15 PM, Yasuo Ohgaki <yohgaki@ohgaki.net> > wrote: > >>> > >>> > >>> Now back to the main topic: > >>> > >>> Please exclude session_serializer_name(), session_gc(), > >>> session_reset(), session_abort() and the "session write short-circuit" > >>> from the 5.6 branch. > >> > >> > >> I removed session_serializer_name() and session_gc() > >> (Although session_gc() is mandatory API, IMO) > >> I like the idea removing INI modifying function in the future release. > >> > >> I don't understand reason why you insist removal of session_reset() > >> and session_abort(). They are just missing API for session module, > >> like session_gc(). > >> > >> There should be API (i.e. function/method or parameters) for distinct > >> operations that user may use. > >> > >> I may agree if you could provide the reason why there should not be > >> these APIs. > > session_abort(), session_reset() are just two functions that somebody > asked about in 2002 via bugs.php.net and nobody responded (11 years > without a single comment) right up until you just assumed it's a good > idea and commited it. If it was "missing API", surely at least 1 user > per year would've requested it. ;) > > I understand that you see them as small, trivial additions, but being > a part of sessions automatically makes them very important and as > such, they should be evaluated collectively, not by a single person. > It does not explain why it is not needed. For example, session_abort() can be used to with error/exception during execution. You may not use, but I would use it to make sure $_SESSION would not contains any unneeded changes when error/exception occurred. > I thought it might be better to explain what new functions do. > > > > session_gc() executes GC without tweaking INIs. > > session_abort() aborts session without writing $_SESSION. There is no way > > achieve w/o it. > > session_reset() re-reads session and re-initializes $_SESSION. There is > no > > way achieve w/o it. (It could be done with session_abort(), though) > > "write short circuit" omits "write" when $_SESSION hasn't change. > > There > is > > no point calling write API and writing to storage for the same data. > > "write short circuit" as I understand it, is an exact copy of the > 'lazy_write' option. This will be addressed in the previously > mentioned RFC that I'll post later today. No. It's not. "lazy_write" does not lock session data while "write short circuit" does. In other words, "lazy_write" changes session behavior, but "write short circuit" does not. i.e. With locked session data, "write short circuit" would not change how session manager works. Regards, -- Yasuo Ohgaki yohgaki@ohgaki.net