On Dec 12, 2013, at 11:59 PM, Mike Orr <sluggos...@gmail.com> wrote:
> > It's certainly worth reviewing how much of Beaker's structure makes > sense in 2013. It's based on the Myghty container API, how important > is that? I've been respecting it because Mike B thought it was good, > but it is odd to retain something that ultimately came from Perl, that > was a side effect of a template engine that went mutant and swelled > into a quasi-framework. well that was before the Catalyst thing and I did at least three major “de-embarassmant” refactorings through Beaker, but yeah. Particularly, the way the Session was bolted onto Beaker’s backends was never really that great. > > So, what we need is something that conforms to Pyramid's Session API, > and has switchable backends. Dogpile provides some of the backends, > and perhaps other backends can be hooked in through it. There’s definitely been interest in putting a Session front-end on dogpile’s backends. Someone may have even said they were working on it, but I haven’t heard anything in awhile on that front. > It doesn't > necessarily need to be multi-framework or Myghty container API > compatible. Middleware has been almost completely replaced by Pyramid > Tweens, and nobody has objected, and non-Pyramid people haven't been > writing parallel middleware. So the same thing can happen with a > session library. it would be super-nice if some kind of “session” thing existed that could be used in any HTTP-like context. that is, don’t hardcode it to tweens necessarily… > like a step backwards, and forcing people into Redis if they want > sessions seems odd too. I can see people just not choosing Pyramid if > its only out-of-the-box session choice was Redis. And it also feels > like a lot of Pyramid developers decided to deemphasize sessions and > Beaker without telling me, so that bothers me to a bit. not sure about others but I haven’t used backend-based Sessions in a long time, I put a few key tokens in a cookie-based Session (for which I am using Beaker, that part of the implementation was written by Ben and has almost no connection to the Myghty part of things) and then any state that is more significant is modeled in the database explicitly, some of it keyed to the session id. This might be why it’s hard to get traction on a backend-based Session system within the pyramid community, we tend to be more formalist about storing significant structures.
signature.asc
Description: Message signed with OpenPGP using GPGMail
