Damian Cugley’s Weblog

I really lost it at work today. Why am I so frustrated? Well, one of the things that upsets me is when stupid software makes simple things hard. For example, when I find myself spending an entire fucking afternoon trying to copy data from point A to point B.

Now it so happens that Unix and the like sorted out this problem a long time ago. On Unix everything is a file, so you move everything using cp and mv. (Or, on Mac OS, everything is an icon in a folder, and you move it with drag & drop.) Between computers you can use ftp or scp. You move a bunch of things using tar. There are a few corner cases, but that essentially covers it.

Not so in Windows world. Every sort of data is stored in a different way. For example: Application settings are in the Registry and require a specially written, and quirky, program to edit by hand. That’s becase we mere users are not supposed to need to edit the Registry by hand. It also follows that the format of Registry-based settings are completely undocumented. So when you do need to fix something, you’re stuffed.

How about Microsoft SQL Server databases? Now, logically you should not need to be copying databases around willy-nilly: the whole point of a DBMS is that it manages your database for you. But no, your colleagues, having been raised on Microsoft Access will expect to be sent database images through the post. So you need to start Microsoft SQL Server Enterprise Manager, and navigate to the database (this taks a while as it searches your LAN for servers), invoke the Backup command from an obscure submenu of the context menu, delete your usual back-up settings, and navigate through some frankly weird dialogue boxes where files are called devices for some reason. You have to be careful where you save the resulting file because the directory it chooses by default is not on a network share and so is inaccessible to you. When you are done you probably ought to try to put the original settings back or you may find nightly back-ups don’t work anymore. At the other end, your recipient has to navigate through Enterprise Manager as before, chosing Restore this time, and locate the file you sent them. They then have to navigate through more insane dialogue boxes where they must enter by hand the file names that their database stores its data in—an obscure implementation detail that is surely the database server’s responsibility, not yours!

Actually there is also something called DTS, which is another way to copy Microsoft SQL Server databses about, but that has various problems. First, it also involes entering weird options into fancy dialogue boxes, each of which instructs DTS to corrupt, fold, spindle, or mutilate your data in a subtly different way (oversubtle transcoding of character sets, or dropping extended properties or whatever). Creating an actual exact clone of your original databse is quite tricky. Second, DTS is specially licensed magic voodoo which means that you cannot, for example, use it to inject data in to a personal-use database running on a portable computer so that you can demonstrate your SQL-Server-based application to clients. Third, while the designers of the system were vaguely aware that certain database objects must be copied before others, they did not have the wit to actually solve the resulting dependency-graph problem, preferring some kludges and workarounds that mean that you don’t usually run in to trouble. Of course when you do run in to trouble, you are stuffed.

All in all, copying databases is one of those tasks that should be simple but which usually leads to half a day annoyingly wasted.

What set me off this afternoon was actually a combination of Microsoft Visual SourceSafe, Visual Studio .Net and IIS. The morning was bad enough: I spent it trying to create a bug-fixing branch of a project I  was working on. (This is where you create a copy of the project as it was before you started working on the next version, so you can fix a bug in the old version. It is supposed to be a straightforward task in a source-code control system.) For some reason the wrong versions of the files kept reasserting themselves and it was a horrible mess. But In the end I got it working and actually made a bug fix release. So far so annoying.

The project in question has two products, both of which I am developing in parallel, and so I had put them all in one .Net ‘Solution’ (=collection of projects). The bug-fix branch only affected one of these products. I decided that keeping them mushed up together in one Solution was wrong and it would make sense to split the Solution in to two parts. Solutions correspond to directories, so this boils down to a desire to move (some of) the project files in to a different directory. With normal programs this is perfectly doable, but one of them was an ASP .Net web application.

ASP .Net web applications take a slightly weird attitude, storing the files on an IIS server, with complicated arrangements to link back to the development environment when the ASPX pages are visited in a web browser. (Setting this up in the first place required that I reinstall the operating system and every bit of software on my desktop computer several times over.) This means that VS .Net hands over control of where the files are to IIS, which in turn means that they are not stored in (subdirectories of) the same directory as the other projects in the same Solution. But I wanted them stored together so that I would be able to copy them around, put them in ZIP archives, store them in Visual SourceSafe, and what-have-you. Oh, and so that they would be on the disc which gets backed up each night.

So I spent the afternoon in increasingly desperate attempts to achive this fairly simple goal, deleting virtual directories and stillborn .Net projects over and over as I tried doing the various steps in different orders and got more and more exasperated. Again and again the computer locked up for minutes at a time as it copied files back and forth between its own disc drives (how hard can that be??). Again and again it popped up message boxes boasting of its uncanny ability to hide the obsolete file names or old URLs in places I could not find them. In the end I stormed off home and spent the evening seething in front of Star Trek: Deep Space Nine. I’ve had a nice relaxing bath now and I guess tomorrow I’ll just have to admit I have no control over where the web server chooses to store my files and try to get Visual SourceSafe to follow its lead. So long as I don’t keep files checked out too long the risk caused by lack of daily back-ups will not be too great. It depresses me that I seem to have so little control over the damn thing.

Ten years ago I wrote my own web server (handling the then-current HTTP/0.9 protocol). I admit it wasn’t a great server, but at least I was in charge of what the computer did, and could fix it if it went wrong. What ASP .Net developers need is an equivalent of the servlet runner that Sun supplied for developers targeting JavaWebServer—a simpler, more easily configured server for developers to run ASP .Net classes with during development. I think Microsoft are reluctant to do this because it promotes the idea that there can be more than one web server, which in turn might make people examine the competition to IIS more.

5 March 2003

Article Archive by Year