> -----Original Message----- > From: Mark Nottingham [mailto:mnot@xxxxxxxxxxxxx] > > Well, that's a bit of a straw-man, isn't it? AIUI 3 *is* already 2 re- > coded into C++. Never mind the question of why that's necessary; > indeed, I think a lot of people's discomfort is centred on the fact > that large parts of 3 have been rewritten and not battle-tested in > wide deployment. Some of my discomfort stems from the fact that from where I sit, development on -3 seems entirely ad-hoc, with no direction; whereas -2 development is entirely focused (of course, with only Adrian really developing it, it's going to be moreso; what I'm talking about is more 'what is being developed' than 'who's developing it' though). I could be talking entirely out of turn here though, as I haven't seen a -3 roadmap. Here's a big thing that absolutely scares the crap out of me as far as moving to -3 goes: Stable Versions: Version First PRE Release Date First STABLE release Date Latest Release Latest Release Date 3.0 Aug 15, 2003 Dec 13, 2007 STABLE1 Dec 13, 2007 2.6 May 29, 2006 Jul 1, 2006 STABLE18 Jan 10, 2008 Now, don't get me wrong here, I understand how development cycles work, and I understand that it was a ground up re-write, but 4 years to get to a stable version that has _LESS_ useful features than a version that was being actively developed, outperforms the hell out of the 'new' version and contains more features overall just doesn't instill confidence in me in terms of moving over to it. Sorry, but that's just the way it is from a business perspective. The second thing, that Mark has touched upon, but the majority of squid developers don't seem to get, is that the big users of squid are businesses. We rest our livelyhoods and userbase upon squid - Yahoo! a hell of a lot, Last.FM a hell of a lot too. It's all well and good saying 'you guys should switch to -3, it's the shit!', however, it would be cowboyish of us to migrate our caching services over to utilizing something that in our own testing doesn't perform as well as what we're already using, and is barely tested at our scale. -2 has proven survivability, we're in our comfort zone using it, we know it works, there are no odd suprises awaiting us in a months time due to untested code. >From a financial aspect, I can say right now that if I upgraded to -3, and in a months time it took our site offline, it'd probably get my ass fired. The truth of it is, as much as you guys tell yourselves that your userbase is people who run one or two cache boxes in their basements to cache their lan internet access, and that there's no money in squid, there are many companies who use it in production as a reverse caching proxy in which it is an integral part of their setup. These users - Mark and myself included -, in volume of cache boxes make up a huge portion of your userbase, and we're the type who shy away from change because it burns us in the long run. We're also the type who are willing to sponsor projects to add features to them so that we don't have to change, and will avoid change like the plague because change means downtime and instability in something that is in all likelyhood already unstable by design. > > Again, parity with -2 isn't enough; why would someone pay for > something they can already get in -2 if it meets their needs? > <snip> > > While I'm in a mood for ruffling feathers (*grin*), it might also help > to have the "core" discussions in public; AIUI there's a separate > mailing list for this, and while having those discussions hidden away > shelters you guys to some degree -- and I appreciate your motivation > for doing so -- it also removes the opportunity for feedback by > interested non-core folks. You might find that some more transparency > improves the process and vitality of the project. I totally agree, and I'll go on to say this, since it seems I too am in a mood for ruffling feathers... I've spoken to Adrian too many times to count on two hands about this whole thing, and if you guys are trying to re-invent the wheel, you may as well stop now. There are so many other projects out there which are doing what squid does now, and in some cases, they do it much better (I'm talking in terms of reverse caching proxy here) - varnishd for instance. You can throw the question back at me 'well, why don't you use that then?' - the answer is above: change. If it comes to the point where I'm forced to change to -3, then I'll be changing to varnishd. It's that simple, and here's why: They're both 'new'. Squid-3 is relatively untested as far as I can tell in the r-c-p arena; I've used it myself on a couple of test machines, and the performance I got out of it was about 1/4 that of the -2 machines, but that's pretty anecdotal. Varnishd is in production at a few decent sized websites, and is being used successfully. Varnishd has all the features I need in it to be a direct replacement for -2. Varnishd was designed from the ground up to be a reverse caching proxy.d And here's the big one: If I have to change, I'll change to something that's better, not something that's the same, or worse. (That's the financial aspect Adrian was alluding to in his original e-mail. Finance isn't always about actual money: If I can get more out of my cache boxes by switching to varnishd (i.e. double my capacity - which testing looks like I can do), then it's more financially viable for me than say -3, where I'll have to add several cache machines to even get back to where I am with -2 today.) If anyone takes personal offense to what I've said above, I apologize, and it wasn't intended. -Tony
