...and Misty_Morning_Dew

[mdlbear]

Yes, I definitely seem to have fallen into a modern-day version of The Programmer and the Elves...

...still a lowly programmer -- and is now stuck for the foreseeable future maintaining this horrid crock of a Fortran program, written by elves! After all, nobody else can understand how it works. It has variables named Shamrock and Rainbow and Misty_Morning_Dew, and some of the most ferocious assembly language subroutines to be found outside the jungles of Borneo.

And the moral of the story is: Never do the impossible. People will expect you to do it forever after.

Well, maybe not quite that bad. But it's a horrid crock of a form and workflow system built on top of a beta version of Java that are both 10 years out of date... Somebody upgraded Java, and it broke.

So I commented out the one place where it was using the old, incompatible KeyStore class, and it's back to crashing a lot instead of every damned time. That's an improvement. Of course, now it's only pretending to create and verify DSA signatures. But since you can't fill in a form without logging in to the server with your Unix password, there was never any real reason for the digital signatures in the first place.

This still leaves the uncomfortable question of why a horrid crock of a research experiment written by two people a full decade ago is still better for our lab's workflow than anything we've been able to turn up in the open source world. Suggestions?

Or is there still some research left in that topic? Hmm.

Comments

[randwolf.livejournal.com]

Open source doesn't respond to users--only developers; it suits the labs needs because it was developed for them. In commercial software, there's marketing. And I used to hate marketing. But at least marketing people listen to the users some; most developers on their own solve math problems.

[phillip2637.livejournal.com]

For the past dozen years I've worked for a variety of companies, developing commercial software products. Before that, I spent a lot more years developing in-house systems. I find that I miss producing solutions to real users' problems rather than the hypothetical ones that marketers and business analysts come up with.

As difficult as it is to extract requirements from a user, it's still more likely to succeed with a real one than with one that only exists in somebody's head.

[mdlbear]

This is nowhere near true, on either side. For products sold to businesses, as opposed to the retail market in boxes, marketing people mostly listen to the IT department's managers, and in some cases the users' managers. Small businesses and workgroups are seriously underserved in this market. Even in the retail market, with a few exceptions, it's less important to listen to the users than to come up with new features that will sell the product to more users.

Successful open source projects are developed by people who are either users themselves or are employed by users. Apache was developed by webmasters. Think our lab's Linux kernel guy (hint, it's not me) isn't responding directly to his users' needs? Think again.

[randwolf.livejournal.com]

But this is when the developer is the user. The problems come when the developer is not the user.

[randwolf.livejournal.com]

Sure. The best marketing--think of Apple--interviews and tests. The worst does exactly what you say. In between there's a range.

[mdlbear]

All that's really required is that the developers be close enough to the user community to be able to communicate with them directly. This is always the case in successful open source projects. Again, our lab's kernel guy, Apache, Eclipse, Mozilla, ...

The developers don't have to be users themselves, though it helps, and the vast majority of users don't have to be developers. A successful open source project has two mailing lists: one primarily for developers (that users can read and even post to if they want), and one primarily for users (that the developers read carefully). It has a bug-tracking system that users can post to and track.

Problems come when the developers have no connection at all with users, and this is much more likely to happen in a closed-source project. If the developers' only contact with users is through a marketing department that has no understanding of the technical issues involved (I've been there, and I know), it's worse than useless. The marketeers keep dropping impossible requirements for new features on the developers, and ignoring real problems that the users are having because fixing them is only going to help a small fraction of the total available market and isn't going to sell more product.

Call the elves....

[hvideo.livejournal.com]

There seems to be a little trouble with your link to THE PROGRAMMER AND THE ELVES. Instead of coming up with the standalone pigsandfishes.com address, it is putting that address at the end of the address for the current page. Looking at it from my friends page it starts out with the http://hvideo....http://pigsandfishes.... and starting from the comments page in your journal it starts out http://mdlbear....http://pigsandfishes....

Once I saw what it was doing it was trivial to delete the FIRST http address and be left with the second (correct) address at pigsandfishes.com - but just clicking on the link wasn't working.

Re: Call the elves....

[mdlbear]

Oops -- thanks. Fixed.

Actually what had happened was that I somehow got part of a tag into the URL through careless editing.