And if so, can it accept SQL text to do complicated queries with rampant joining, or does it handle that sort of thing fairly well by itself?

I’m using it on a small project at the moment. It allows for arbitrary SQL or even substitution of stored procedures for SQL statements if needs be at the object level, while hiding that stuff at the pointy end.

I’m pretty sure it does generate SQL down at the connection level but it’s hidden from you unless you feel the need to tinker. I would anticipate that the SQL server only stuff will disappear very quickly as other database vendors ramp up the amount of meta data their drivers supply.

As long as your database is fairly standard, and you are happy to be co-opted by the MS techniques for concurrency management based around timestamps, it’s about as good as O-R mappers come. You have to design your database more-or-less upfront though and be prepared to stick to the tried and true techniques of foreign keys and sensible indexing to make it perform and give it enough meta data to draw inference about the object relationships. It’s happy to do “lazy instantiation” for objects you might not need straight away, a lovely feature that the writers of these things often forget to do.

The bad? You still can’t display arbitrary properties from objects a couple of foreign keys deep on a grid without implementing your own custom property, but arguably you shouldn’t be doing that anyway. Like most of these things, the object relationship to collections is clunky, but I’ve never seen it handled very well anywhere. You’re also still stuck with the age old object versioning problem and better be prepared to smack your developers up if they don’t write change scripts when deploying new versions of the software. I’ll bet they all forget that step.

i.e. overall I think it’s good (good enough at least), and having written a few of the stupid things over the years am heartily relieved by it’s presence for all the boring sh*t I hate doing.

Whether it generates SQL is not something I recall. I suspect it’s tightly bound to SQL Server so perhaps it’s relying on a different representation.

I don’t think it’s a pre-processor though. I believe it compiles directly to the bytecode and uses all the typing hooha C# lends.

LINQ is an honest-to-god innovation from Microsoft, if youre prepared to ignore the fact that you could do it in Lisp to broadly the same effect.

Or dozens of 4GL languages from the ancient past where you invoked a pre-processor to parse your embedded sql-like-language. It’s not a new technique, but it is very nicely done. Which means it will get used a lot, I think.

So maybe something like a Tesco/Linux OS isn’t that far fetched.

LINQ is an honest-to-god innovation from Microsoft, if you’re prepared to ignore the fact that you could do it in Lisp to broadly the same effect.

I guess we’ll need to agree to disagree. I give you the benefit of personal experience — I was a kid when Unix became a trademark you could apply for. In any case you have to agree that the next few years are going to be interesting!

I think you miss my point, because you tell me the “Ubuntu experience” but the question was originally about “Linux”. Ubuntu is just one flavour of linux. So you’ve sort of skipped at least 3 steps until you get to “the easy bit”, and even then your answer is contentious (see below). This is one of the big reasons why mass Linux adoption is so hard — the great thing about Linux is that there’s so many to chose from! (and they are all different).

I was thinking on the way to work this morning about how hard it is to get software into a standard Linux repository. Since I write commercial software (the shame!) there is no hope that my software could ever end up in a large Linux repository. Ironically, this means that the repositories themselves are not free — they are actually gated communities, strongly biased against commercial use. I respect their right to operate this way, but it’s not ideal for end-users. An opposite example is the soon-to-open iTunes App store, which does in fact provide a repository (and a commercial mechanism).

Replying to your other comments:

I do realise that Leopard is a UNIX; that was actually my point. If it doesn’t already do so, Apple is going to dominate the market for desktop Unix, at the expense of Linux. This is going to hurt Linux’s position vis a vis Windows. In my opinion, the question is no longer can Linux win, but can Apple, which is sort of the point of all of my postings.

Regarding finding OSX in stores, I think you miss my point again. I’m not trying to be harsh, but you said that Windows doesn’t have a central repository, and I said that it does, and that they are called “shops” and “online stores”. Although it’s not obvious, they are actually more powerful and less closed than Debian or Red Hat repositories.

But Windows is fiddly. Which version of Windows? Do you have MSI 2.0 2.1 3.1 installed? .NET 1.0? I require .NET 3.5.2! DirectX 7, 8, 9 or 10? Have you registered the software? Whats the key? Too many attempts. Install, Deny or Allow? Uh oh, DLL conflict detected in pre-.NET code. Guess youre SOL.

This is a dangerous comment to make when comparing Windows and Linux. I have been in RPM hell, and APT has become much less friendly since Etch. Now I’ve also heard about DLL hell, of course, but I’ve also spent far too many days compiling stuff on Linux or trusting backports from people I hardly know, because I can’t find a compatible RPM or DEB. So I’d say that at best, these two systems are a tie.

But I think we are actually in broad agreement. I’d love a release of Debian with an OpenSolaris kernel – I am hanging out to be able to use ZFS for my phat replicated NFS servers. I agree that OpenSolaris (the OS) is probably horrible, but I wouldn’t agree that RPM, YUM or APT are perfect packaging systems; and I think Murdoch’s entrance into Sun is a very good thing for Sun (not so much for Linux).

At the moment OSX is my operating system of choice, and I find it difficult to imagine a time when that won’t be true. It’s easy to use, it’s Unix, I can install gawk (via Portage), it looks pretty and it works amazingly well on my new MacBook Air, which is the best laptop I’ve ever owned (way better, subjectively, than my supposedly more powerful MBP). Apple are kicking a huge number of goals at the moment, IMO they are already beating Linux and I think they’ll eventually supplant MS on the desktop; but it’s a 10 year job.

All of the preceding technical discussions about (eg) MySQL v. Postgresql, dev tools, etc really miss the point I was making, which is that Linux is a mess of egos and that this mess is a replication of the old days, when Windows beat the crap out of Unix. Nothing has changed — except that (arguably) Windows has gotten better and Linux has gone nowhere — and so there is no reason to think that Linux will beat Windows.

Cheers
Mark

Particularly MS SQL server. Id probably killwell, maybe commit grievous bodily harm to get Oracle into my dev stack.

OK – SQL server is the red headed step child of database servers, I can agree with that :)

It’s still better than MySQL though, which is primordial ooze in comparison.

So with that background in place, I agree with everything you just said. Particularly the bit about weeping with joy. Over the years I’ve gone a long way to implementing LINQ myself because if it was getting to wrist cutting time whenever I had to write another class to dump and retrieve data.

I agree right up until the last sentence. The IDE is all right even if it is memory hungry, but the rest of the dev stack sucks arse. Particularly MS SQL server. I’d probably kill…well, maybe commit grievous bodily harm to get Oracle into my dev stack.

I think a lot of people over-estimate the nerdliness of most corporate programmers. The vast majority of them are business domain specialists who program to specific problems, not uber-nerds who have paroxysms of delight over the “correctness” of their chosen tool or delight in wars over which text editor is the best. They need something that’s easy to learn and everybody else knows, so they don’t have to solve 100% of their problems by themselves and other programmers who know the tools are easy to find. They are not “great” technical programmers, but they don’t have to be.

The LINQ stuff is better than Active Record, not because it’s technically better, but because Microsoft wrote it and integrated it properly into the development environment. Any lowly business domain programmer will be weeping buckets of joyful tears as they don’t have to write yet another class that assigns the contents of their objects to a database table with a tools they know will be around forever and available everywhere. They literally drag and drop their tables from the database onto the design surface and the tools do the rest.

In one sense, yeah, LINQ is just another O-R mapping tool, they’ve been around for ages. The killer difference is that Microsoft is giving it away, you don’t have to buy it, so it becomes universal amongst that class of programmer whose entire raison d’etre is mapping endless amounts of business classes to database tables. That’s a massive win, it’s such a boring job and now it’s basically automated so they can devote more time to the logic of the application.

For some classes of problem (say networking issues, video streaming, web page serving), Linux is absolutely a better choice. For boring business applications where the database is the computer, MS dev environments kick everything else all over the shop including Java.

I’m also not so sure about how much of a killer app LINQ is going to be. It’s pretty much just an Active Record implementation, no? It’s not as if they couldn’t use (the ruby implementation) ActiveRecord or the implementation that drives Django; and I bet there’s a Java implementation somewhere as well.

Visual Studio (Microsofts development environment) is their last killer application. Ever since Visual Basic 6 killed every other proprietary and free GUI based programming environment, Microsoft have had a lock on the desktop. For starters, the 3rd party applications in vertical markets are massive and pretty much all Microsoft. It looked like they would all shift to the web 10 years ago, but it never happened. Even inside big companies, there are probably billions of dollars of applications written in some variant of a Microsoft language (from excel macros to full blown C++ and C# applications and everything in between).

None of that stuff runs on Linux easily (advances in Wine notwithstanding).

Before I get jumped on, I know how much of a crock VB6 was, but everything else was much worse for a workaday programmer who isn’t interested too much in uber-technical stuff. Visual Studio 2008 is a really nifty tool and fixes much of the broken-ness of the earlier MS languages, and is probably set to extend that Microsoft desktop dominance for another 5-10 years. Why? Linq 2 sql. It’s a departmental programmers wet dream, and once they start working with it, they’ll quit the job before switching tools.

Earlier someone said that wasn’t even remotely close to usable for the average user. I disagree. My experience is that for the average user the difference between Linux and Windows is UI and even then just the look-and-feel component. For anyone doing email, web and small documents it is more or less the same as Windows.

The only real problem these people are likely to run into are printers and I fully acknowledge that when you come home with a winprinter, the kind where the manufacturer has outsourced a bunch of the processing to the OS, and you try and plug that into a linux box, you are in for a world of misery.

But generally I find it is the unusual users who run into problems, people trying to do more esoteric stuff like dual-head or voice command. People like me who, no matter what they want to do, can find something that will give them just enough rope to hang themselves. In the windows world you’d either pay out a bunch of money for a completed solution or – in many cases – there would be no rope at all.

People who are willing to experiment with another OS are not average people.

The ribbon is a sad story of how familiarity defeated design. Its a brilliant design backed up by scads of research, testing and hard data, but its unfamiliar to people whove gotten used to the quirky, ad-hoc arrangements of functionality in every version of Office up to 2003.

You know what was a good design? Backed by about five thousand years of research and in-field testing? Written Language. Damn awesome invention. You have these things called words, and each word is assigned to a particular meaning and people learn the words. A typical human can remember several hundred thousand words and generally achieve instant recall when they want to say something. A modern computer can search through tens of millions of words (even mapping against synonyms and partial matches) faster than you can blink.

Think about the most successful application of recent years — the google search system. What is it? It’s a command line interface. You type words in a box and it goes and gets a response to those words. After all this time experimenting with menus and GUI and point and click, we are no reinventing what we already had. By all means have a go at redesigning the google API to work on purely point-and-click (don’t bother building it, just go through the theoretical experiment with a few sketches on paper) see what a complex mess comes out as the result.

Agreed on MySQL, though theres some degree of transaction support in 5.1 and up.

Pick the right tool for the job. PostgreSQL has supported transactions since forever, plus ultra complex index options and user definable functions and all that heavyweight stuff — for jobs where you need it. MySQL was designed to be lightweight, fast, easy to use and simple. For many applications that’s a perfect match. If you really want heavy, then buy Oracle (and might as well buy the whole Linux from them too so you get bottom to top support), or even but IBM’s old DB2 which also runs on Linux.

If you want to load up several million rows then scan them linearly to produce summary information (NB: index is useless here) then MySQL will win every time because scanning bulk rows is the thing that it does really really well.

John, call me when Excel supports more than 65,000 row because thats its maximum also. I think youre fishing pretty hard for a bunch of nitpicks, some of which – like this one – are just wrong.

None of them support 65000 rows, none of them support 50000 or even 20000. Sure, you can load 65000 rows into the sheet, but try doing calculations, try anything even remotely complex and the thing just grinds to a halt. Memory usage blows out to over 1G and believe me, you better enjoy getting cups of tea, because that’s all you are going to be doing. Now do the same calculation in MySQL and it will say, “What? Only 65000 rows? Are you kidding me?” Even grunging perl over a massive text file will grind through rows faster than any spreadsheet you care to name.

Spreadsheets are great for the presentation of results data to management, with a bit of a cheesy graph in one corner, big headings around the place, and maybe a handful of tweakable options to try some simple scenario analysis. You can do neat stuff in maybe a dozen screens worth of spreadsheet but that’s about where it ends, for big jobs, use an appropriate tool.

I respectfully disagree with your analysis of NT being pretty much state of the art when it came out; it was largely based on VMS, which had been around for donkeys.

Sure, but VMS was a pretty good design to start with /subjective-opinion. But the original Cutler architecture and the actual thing post-NT4 are two quite different things.

like Linux, was fragmented between the HP, DG, Sun, SGI and other big players at the time.

Contrariwise, Linux has pretty much forced convergence on a single kernel. Most of the big Unix vendors either died or gave up. Only Solaris is still really going on its own, whereas technology has migrated into Linux from SGI, IBM, HP and DEC/Compaq.

Oh, and OS X is technically a Unix™ as of Leopard.

* google (familiar) or search your favourite online store;
* click to download (familiar);
* (double-) click to install (or for OSX, drag to install).

In practice you won’t find OS X software in stores; I certainly haven’t. OS X installation is definitely a doddle.

But Windows is fiddly. Which version of Windows? Do you have MSI 2.0 2.1 3.1 installed? .NET 1.0? I require .NET 3.5.2! DirectX 7, 8, 9 or 10? Have you registered the software? What’s the key? Too many attempts. Install, Deny or Allow? Uh oh, DLL conflict detected in pre-.NET code. Guess you’re SOL.

The Ubuntu experience:

* Load Synapse.
* Select package.
* Install.

And it categorises the software for you when it installs it. Even OS X hasn’t gotten to that yet.

I agree on Mac OS X — I’ve used a Mac for about 2 years and you couldn’t make me leave it for quids. But Linux definitely rules the free server roost. I’d like to try Solaris, but — ironically — it is less usable and accessible to noobs than Linux. Plus they’ve decided to invent another packaging system.

Still. If Debian ever do officially adopt OpenSolaris as one of their adopted kernels, Ubuntu will see my dust on the server :)

Linux, due in part to its roots, tries to cater to developers, novice end-users, terminal geeks, X windows geeks, Unix geeks, corporations and the free software movement. Many of these groups are antagonistic to one another.

And this is the core of the issue – “Linux” is a catch-all, the term means something different to everyone. A Field Guide to Free Software Supporters is a good overview of the various positions. So does getting people to use free software on non-free (Windows/OS X) help Linux beat Windows? I’d argue yes, as it makes it easier for people to move later on.

The other thing is the Microsoft monopoly has meant people expect there to be one platform uber alles which will beat all the other platforms out of the market. Which is ridiculous. Different software is appropriate for different markets. Linux is fortunate in that it can be adapted to many markets, but this doesn’t mean everyone will want to use it for everything. The growth of OS X is also good from this perspective – programmers will have to stop assuming everyone runs Windows.

To get back to the original question, Linux won’t beat Windows on the desktop, but that doesn’t matter because the desktop is not a growth area. Mobile internet devices are where it’s at, and Linux is definitely mainstream there. The desktop is a hard market to crack because of the network effects of the Windows monopoly, for business it’s the random ISV app that is essential for them only runs on Windows, for the home it’s gaming. I run a Linux desktop at home and gaming is acceptable thanks to Wine, but I’m a geek and fully understand why all my friends run Windows XP – it’s the platform for games. For all other everyday needs free software is pretty good. Note that Photoshop, MYOB, SPSS etc. aren’t everyday software, they come under “random ISV app that’s essential” since some people can’t live without them, others don’t care. Tax software – eTax isn’t available on OS X either.

Drivers: JohnZ said “every Dell laptop I bought over the last few years had at least one major problem.” These days Dell is pushing for hardware with good Linux driver support, or you could buy from ZaReason who advertise working suspend/resume on some models. Intel pushes their drivers into X pretty quickly, however the lag to when they end up in distributions is still a problem. Small companies can get the Linux Driver Project to write drivers for them.

The open source stack is a mess of different and competing egos, nobody is in charge and nobody really talks to each other. Linux distributions are the Unix wars all over again; but this time the wars are being fought without very large companies spending lots of money; its doomed, I tells ya!

One of Sun’s selling points for Solaris is that it is one stack managed by one company, so things do work together, or at least you have one place where you can go and pay money to get things fixed (since they participate in a lot of upstream free software communities these days).

Exchange support is being worked on by the OpenChange project, which will be integrated into Evolution eventually. Windows costs about $100 bundled, not $30.

I was actually thinking of typos more than my (less than spectacular) spelling prowess.

I respectfully disagree with your analysis of NT being “pretty much state of the art” when it came out; it was largely based on VMS, which had been around for donkey’s. My own experience with it was in the mid 90′s, when I moved from Linux to NT 3.51 on my desktop — and moved back very promptly to Linux, because I had to reboot to add a virtual IP address to an interface (something I did a lot of in the 90′s).

I think that the Novell admins at the time also felt this way. We all understood that NT was an immature operating system missing many important features, and it ultimately owed it’s success to non-technical factors. In my opinion, NT succeeded in large part because it was a single coherent system; Microsoft’s increasing marketing power was also significant, especially against a poorly performing Novell. Unix was easily a better server OS than NT at the time (and IMO still is) but, like Linux, was fragmented between the HP, DG, Sun, SGI and other big players at the time.

On the other hand, Windows doesnt ship with thousands of contributed programs integrated into the overall package. Microsoft dont offer centralised, integreated repositories of software. There are places where it balances out.

I disagree, both with the idea that MS doesn’t offer “centralised, integrated repositories of software”, and also with the implied notion that such a repository is actually an advantage of Linux.

In the first, while MS does not offer a centralised repository of software in the same technical sense that most Linux distributions do, any local Dick Smith, HT or Software City certainly does offer a centralised repository. In most cases they are far easier to search than “rpmfind” or “packages.debian.org”, and there is actually a better selection.

To illustrate my second point, to download and install software for either Windows or OSX you simply:

* google (familiar) or search your favourite online store;
* click to download (familiar);
* (double-) click to install (or for OSX, drag to install).

To download and install software for Linux from the GUI, you

* umm… is it Red Hat or SUSE or Debian or Slackware or Gentoo?
* is it an old version of Red Hat or a more recent version?
* is it GNOME or KDE or text based?
* If it’s Debian, have you configured your /etc/apt/sources.list properly?
* Have you updated the repository cache? dselect AND apt caches?
* Did you know that installing “vim” will require the installation of another 17 packages? Really? Because, it’s like, important, right? PROCEED Y/N?

I don’t want to give the wrong impression, since I love unix, it has been my development environment of choice for almost 20 years. But Linux is clearly repeating the same mistakes as Unix did.

Unix is awesome, but a great desktop is essential, and both of those reasons are why I bought a Mac.

Why do you say you need a ‘spell-checker’? What on earth would you do with one? Surely you went to school before they stopped teaching English? Useless, and unnecessary.

I would agree, though, that a dictionary and a thesaurus may be worthwhile features of a word processor.

1. One of the things that the early Macs were good at (and which was carried over to Word for the Mac, the first application MS ever wrote) was teaching you nmenonics – control + N for ‘nest paragraph’ for instance. So it was really a dual system – visual to learn and to convey metaphors but that taught you a second track if you wanted to follow it – keystrokes for familiar operations.

2. From memory MS took this into Word for Windows, but they eventually took the GUI literally to mean that pointing and clicking was always better – which of course it’s not.

3. The ribbon is a logical extension of this. I don’t want to click a button on the ribbon to do lots of things, I want to use keystrokes, and the ribbon keeps you further away from learning them than ever.

4. I can only speak for myself I guess, but it’s just possible that the ribbon is a better interface to learn on – though I doubt that. Because at least for a sophisticated user, you need to look new things up. And trying to guess what an icon might look like, rather than racing down a pull down menu and checking out the words as they go – well we still use that technology in libraries when we’re trying to look stuff up – we look for ‘G’ if we want to read about governments. We don’t look for a picture of a government, and we wouldn’t know what it would look like.

So I think whatever their research showed, it was misguided from the start. And insufficiently humble. It’s a big change, so why not at least at the outset assume you might be wrong and give people a choice of interface?