New Product From 280North: Atlas

To nip this one in the bud: Atlas doesn't generate code. It serializes the objects that represent the "views" and other objects. This has a number of advantages over the code generation method.

Also, I think Atlas allows for the sort of abstractions you're looking for. It comes with a bunch of built in widgets which are great, certainly for prototypes and simple UIs, but even fairly advanced ones too. But if they're not good enough, or you have some custom widget/layout you can easily add it to Atlas as a plugin and treat it as if it were a built in component.

But I'm certainly interested in hearing more about the problems people have with these sort of tools.

Apologies in advance if this is rambling.

I'm an IT contractor (never done web for money). My opinion was formulated based on my experience managing an IT division over 5 years or so and also doing various PM and architecture roles.

You can more-or-less validate the first 50% of my claim by giving a bunch of computer literate people some brand new boxed PCs or Macs, and an internet connection and telling them to go for it and set up their environment. They'll be able to get things up and running, and if they have a little bit of nous they will usually come up with something workable.

Of course, it won't be engineered. I did some work for a fairly big manufacturer about 10 years ago that opened a small office on the east coast of Australia. The small office got absolutely no IT support as it was off the radar, and because of the corporate structure Finance left them alone too.

They managed to brew up the essentials of what they needed using Access for sales/marketing, MYOB for the books and corporate webmail and hotmail. When IT found out, they went spare and "fixed" everything. The office suffered in both performance and morale because instead of their terrible (but working) homebuilt CRM and accounting systems they had to use a set of ancient AS/400 apps that were about 1000ms away (no problem for interactivity because they were on 5250 screens, but very slow turnaround). The PCs were useless too, IT made them log in to Netware and do file/print over an 8kbps CIR Frame Relay connection.

Government (and to a lesser extent business) exposing and unifying APIs and SAAS probably means that in 2009 situations like the one I messily chronicled above will scale, because the IT development and infrastructure component just isn't required. Someone somewhere else who knows all about scaling, security, DRP and backup has done it for them and is happy to charge a per-user-per-month fee.

If this happened now, they might have used a 37Signals product and Salesforce for example (no idea about the ERP side).

My current contract is a guerilla finance project to build some analysis tools for SOX compliance that they haven't been able to get corporate IT to deliver on for 3 years. We built a small app in Python with a web front-end (Django). It was quick, it worked, and it's a manifestation of people realising that IT has become commoditised and that IT doesn't have to be hard. I guess IT wanted to do something like build some ABAP programs in SAP using consultants that charge £800 a day on a 2 year timeline.

There's a lot of law firms, design offices and small practices that just go out and buy some macs and never even call in IT help. It's going to happen and keep happening, and it's difficult to see how the shift away from corporate IT could happen if you're focussed on changing the VAT rates in FORTRAN, applying PTFs to the iSeries, or keeping the active directory backed up. Shifts aren't usually obvious when you're in the thick of it.

I don't think the quickest typists in the typing pool ever thought that the managing director would type his own memos 30 years ago, either.

I think you're not aware of the Cocoa development culture that spawned Atlas/Cappucino.

People in this culture don't tend to build the kind of endless, mindless forms you're talking about. They build real apps.

And they don't generate code--they generate linked objects (freeze-dried in .nib/.xib and now .cib? files) that are unfrozen at runtime.

If you had to generate a whole bunch of mindless forms, you're right--metaprogramming solutions would be a big win.

"You can't build a UI builder with a UI builder". :-)

I like your point about users "reacting" to an limited range of choices, explicitly stated (and surely using domain specific terminology). It's a grammar you don't have to internalize, with libraries you don't have to know about, or know the search terms to find term. It's very "discoverable", and the limited and clear range of choices minimizes the information content (increasing the probability of obtaining the desired outcome). I'm sure it would also prune the tree by omitting choice combinations that don't make sense.

I think people don't like the system guessing for you, like Office's Clippy. Maybe it's OK if the user asks for the guess - like getting a list of search results; or vim/bash completion; or suggestions (google, Stackoverflow)