Mach-II - Why Bother?!

I refer you to the Mach-II FAQ for the "official" answer to your question:

http://www.maestropublishing.com/downloads/Mach-II...

Now that we're past that, here's my two cents:

It's a matter of which you are most comfortable with. Does that mean that Mach-II needs to "die a dignified death"? Hell no. Mach-II is far from obtuse and quite easy to understand once you master the principles behind it and, in turn, the principles behind MVC frameworks in general.

I, personally, came to CF from C++. OO concepts and MVC frameworks are nothing new to me. When CFCs came along, my reaction was "About damn time." Other people who started out in procedural languages (CF 5 and previous included) seem to have a MUCH harder time grasping both OO concepts and the MVC concept. This is part of a greater knowledge gap in the CF community that is the topic for another day.

If you are asking this question, then you should also be asking, "Why use a framework at all?". You are not just questioning if Mach-II should be used, you are questioning if a framework should be used at all. It's all a matter of choosing the tool you are most comfortable with for the job. Am I going to tell you Mach-II is better than Fusebox (well, maybe that one) or Model-Glue? It depends on the comfort level of your programmers with the concepts and the tool that works best for your project. Regardless of which framework you use, your code will be much easier to maintain throughout the life of your application.

Peter, I am choosing not to respond directly to your post because others are using it to take this debate to an utterly silly and petty level. Nothing against you, I know you are just tweaking to cause debate, but apparently there are some who have difficulty distinguishing the subtleties.

Just to add to this Peter, there are sometimes non-techincal reasons a framework is adopted - depending on your teams experience etc. but i'm sure you know that. One of the reasons I opted for using Mach-ii was simply maturity and the fact that it was proven in the wild (@Macromedia) - Not saying MG isn't but I would be interested to hear people's experience using MG on large sites/systems. I have'nt used MG in anger so I couldn't compare the two. I have noticed there seems to be more people willing to use/adopt MG as opposed to Mach-ii - I think this is partly due to the *percieved* learning curve. From what I remember at Sean's framework preso (last year) he pointed out that fact that Mach-ii is good for websites with highly dynamic flows not sure how MG stacks up in this department.

@Brian - understood. But I'll still bug you for input on IM as it is a genuine question behind the pushy title and I really want to nail down some meaningful heuristics.

@Sean - Yeah, anyone who hasn't downloaded all of your "software" links on top right of your blog should do so now (as long as they don't ask you to justify your performance comments from those one or two slides in the Duck Typing presentation :->). The last time I read that presentation (many months ago), I was pretty new to CF OO so the subtleties between Mach-II and MG eluded me. Rereading, it seems the main highlighted point is the distinction between dynamic controller flows in Mach-II and static flows (with redirects) in MG. Anyone played with both frameworks enough to give some specific examples of the gotchas that each approach provides when you start to build larger real world apps?

@Kola - Probably the best comment. A while ago Joel Spolsky said something similar about languages. The best one is the one your team is familiar with - not the one that happens to have closures or other fancy language features which are nice but not as important as the right team in a language they are comfortable with.

For a while I was wondering if Mach-II was dying off as it was, er, hmmm, REALLY stable (no changes at all :->). Obviously that has changed in the last few months and while I think the FB vs. MG or Mach-II heuristics are pretty clear (MG or Mach-II REQUIRE OO skills so FB is a much smarter choice for procedural or transisioning teams while still having a place for OO development teams who are comfortable with it), but I'd like to tease out any more subtle distinctions from anyone with deep knowledge of both Mach-II and MG because now that both MG and Mach-II are in the wild and supported (Mach-II has a longer track record, but MG 1.1 seems pretty stable from what I can tell) the question of which to use for an organization that doesn have deep experience with either is certainly of interest.

Also Sean I know you are the king of "lets just take this application and port it from framework X to framework Y". Any general hints or heuristics about how to think about your architecture in a way that it would make it quicker and easier to port from Mach-II to MG or vica versa? I know it shouldn't affect your model much and I'm guessing some simple mappings would allow you to port your views fairly quickly, but what about controller logic for a larger site. Any thoughts?!

Peter, I know you're just trying to stir the pot (again), but calling Mach-II obtuse illustrates a lack of knowledge (I'll stop short of saying that it illustrates an ignorance) of Mach-II. Read the Quick Start and some of the intro docs, and then why don't *you* explain to *me* why you think it's obtuse. Yes, there's a barrier to entry with the necessary knowledge of OO principles, but that barrier of entry isn't the fault of Mach-II, and I would argue that if you want to build MG apps you face that same requirement.

I'll probably refrain from further comment as well because, like Brian, I'm getting tired of these sorts of debates. If you design your application correctly you can easily switch from M2 to MG (or vice-versa) with very little trouble, so the decision between frameworks really isn't all that important. Try them both out and choose the one that you think works better for you.

In short, many of these discussions are silly, unnecessarily contentious, and I'll spend my time building apps and working on my framework of choice instead of getting sucked into what in my opinion are rather fruitless debates.

Thanks for the follow-up Peter. I suppose I find the discussion contentious for a few reasons, although you clarify quite a bit in your most recent comment. First, the title of this post itself doesn't set a tone in which rational discussion will likely take place. So when I said I was getting tired of discussing these types of issues, it's at the high level ("Which one is better?" and "Which one should I use?") that I find things rather nonsensical. I don't mind at all having intelligent dicussions with people who have taken the time to become informed about each framework.

It reminds me of a graph I saw a while back concerning open source software. The number of people who will spend endless hours blathering on about why open source is good or bad is huge. The number of people actually putting their money where their mouth is and writing OSS or contributing to open source projects is miniscule. The parallel here in my mind is I think there's a huge amount of damage discussions like this can do to people who haven't tried either framework when they see "Mach-II: Why Bother?" and don't go any further, or listen to a moew vocal side of an argument, no matter how incorrect it is. That's the sort of stuff I simply don't have time for.

I appreciate the question concerning Mach-II's brand, although if MG's brand (given the name and discussion I'm assuming this is correct) is "Unity," meaning integration with CS and Reactor, then as we've said, Mach-II's brand would be that we are going to continue to focus on creating an excellent, highly proven MVC framework that allows easy integration with any other framework you want to use if you choose to do so.

Branding, however, is probably too high level to get to your point concerning heuristics, and more important your distinction between Toyota and Volvo. I would argue the distinction between M2 and MG is probably more the distinction between Honda and Toyota. By that I mean you really have to look at the finer-level details if you truly want to see the difference between the two. So at some level the burden is on indiviudal developers to spend some time looking at each before intelligent discussions can be made.

Remember that although M2 has been around for a while, it's still a pretty young framework, and it's only with the latest MG that the distinctions between the two have become more stark. We have a lot of ideas for new M2 features and have a bit of a roadmap sketched out in our heads as to where we want things to go, so the "branding" will become more apparent over the coming months.

This gets back to the OSS discussion I made earlier, however. If people think M2 is lacking features or it's unnecessarily (key word in this sentence!) difficult, we want to hear about that, and better still, get involved with the project(s) if you think you can help. I just get the impression that people spend an awful lot of time talking about such high-level and ill-informed things as to make them rather pointless.

Finally, I think it's interesting that the people who use (or just like to participate in the discussions about) the frameworks seem to have more divided views than the authors and contributors to the frameworks themselves do. I think Joe's a great guy, I respect him, and I agree with a great deal of what he said in his latest blog post about all of this. M2 and MG were amazingly similar for quite some time, so right now in the evolutionary path it might seem difficult to distinguish. For me, I find M2 more straight-forward to use, therefore I prefer it. Maybe it's my Java background, maybe it's just how I think, who knows.

Point is we have choices and we're all maturing as a community far more than we would have had Ben, Hal, Joe, and everyone else who's contributed to these two frameworks over the years not taken the time to push the envelope in CF. My only hope is that we can continue to grow as a community and have intelligent, informed discussions about these issues rather than inflammatory ones that lack substance.

Hi Matt,

Many thanks for the detailed, thoughtful response. Firstly, your point about the title (and some previous titles) is well taken. I am still a complete n00b at blogging and am not sure how to handle titles. The truth is that the contentious titles get WAY more comments and I only learn when I'm listening (to comments), not speaking (writing posts), so I'm probably going to keep them up but I will see . . .

While I took a little heat for Frameworks Sucks, I also learnt from some great posts by Hal, Sean and others. While this title is probably also not popular with some good friends, I think it helped (along with Petes lawnmower response to my last such posting) to get a great comment and then post from Joe and now a great and thoughtful comment from yourself.

The only thing I promise is that I'll try to follow up with some more moderate and thoughtful posts as I learn more about the heuristics people use when selecting a framework for a project. Of course, the only problem is that nobody will read them as the titles will be too boring :-> (If you look at my blog you'll see the majority of the titles are fairly sedate or yawn inducing!!!)

I tend to agree that "branding" is a little optimistic. It would be easy to brand MG as "the quick fix" and Mach-II as "the enterprise choice", but as Joe pointed out, MG isn't just the scaffolding, and as you pointed out, Mach-II isn't that hard! Each framework supports a wide range of use cases which defy easy branding (and given you don't have to sell them in a 30 second slot, that is perfectly OK).

I agree that it is the responsibility of developers to put the time into choosing the right framework. I also know many smart people have already done that and I can't beieve that none of them noticed that certain classes of problems (other than just dynamic vs. static control flows) might more easily be solved by the collection of metaphors and abstractions selected by one of the frameworks. The challenge is that many people don't examine their heuristics, using general "this just feels the right way", and it often takes some prodding to elicit the remarkably simple rules we all actually use to make sense of the world and make decent decisions quickly. So I'm just off heuristic hunting right now to shorten my learning curve.

I also look forward to informed debate. The only point I would make is that I feel some of inflammatory headlines (followed with more balanced discussion) have actually done a very good job of teasing out great distinctions from very smart and busy people like yourself, pete, joe and many others and I feel that has a place in helping the community to continue to evolve.

Here is to contentious titles and reasoned debate. Let's see if the two can co-exist! (If not, the titles will go).

Best Wishes,
Peter

Peter,

Please send me some sort of source material on where I said that ModelGlue "is the one true way". I said in a comment on Peter F's blog that MG runs faster, is easier to learn and to work with, and does the same job as M2 with less clutter and fewer objects.

No where did I say M2 has no value, should go away, "die a dignified death" or is in any way unworthy... nor did I say that MG was the "one true way". In the context of that article, was I was pointing out was that his comparison was flawed and his arguments were absurd to the extreme and if he REALLY wanted to go toe-to-toe with MG he was going to have a hard time on technical ground or performance...

Aside from that, Joe is one of my very good friends and I have a LOT more experience with MG than I do with M2, so I mildly resent the assertion that I'm promoting the "1TW" approach to things here. If you're going to make the assertion, can you please provide some sort of citations?

And Matt,

Last exchange with Jared convinces me. I'VE had enough of these framework posts. I'm just going to get back to application generation and design patterns and if I happen to come across any good heuristics for selecting a given framework for a given use case I'll post it with a nice sedate title.

I'll keep the controversial titles coming, but not on frameworks. Not worth the animus (which was soo not what I was aiming for). Especially as the people I'm pissing off (on every side of the debate) are the ones I've learnt the most from and respect the most.

Best Wishes,
Peter

Hey Jared,

No harm done - for sure! I guess it was a little optimistic. I was hoping to get Joe to highlight where he felt his design decisions would be less good than M2 for certain classes of use case (the trade offs we all make when deciding what variability to support and how to implement it) and Peter and Matt to highlight the weaknesses in M2 for certain other use cases so that over tea and crumpets we could all suggest that while learning both frameworks was ideal, for application x, we all agreed that MG would have benefits and for application y, we all agreed M2 would be a better choice.

Oh well, at least no blood was spilled :->

Best Wishes,
Peter

@PB, one of the problems here is that the main differences between MG and Mach-II is mostly semantical and thus subjective. To me, it's impossible to objectivefy a subjective subject. I'd be more than happy to quantify my thoughts on a rating scale (like 1-5) -- however the scale is only quantifying an subjective subject. That's why the answer of "it depends" works well.

Really I'm trying here and I'm sorry that people felt that my post was taking personal digs at MG or Joe or anybody else that I'm not aware of. Honestly, the whole "framework debate" needs a rest and I'm personally tired of debating the same stuff over and over.

Ok, time to get back to work people! Less debating and more framework coding otherwise we're never get new versions out!