Older blog entries for niksilver (starting at number 16)

Read an excellent article on the O'Reilly Network about How to build a simple Web Service using MS's ASP.NET. It's easy for a Java person like me to skip through, and it highlights several really interesting things:

• MS's IDE makes it really easy to make a method available via SOAP, including an informational Web page about what methods are available, plus simple forms to access them. A lesson to us all in ease of use and how to demonstrate it.
• Web Services are are staggeringly unambitious --- it's just remote procedure calls via XML. Possibly the most uninteresting so-called revolution since XML!

Web services (by which I mean SOAP, plus UDDI which is a directory service) is helpful in getting businesses to work together, but it doesn't make it any easier to create the services (you still have all the uptime problems) and it only makes possible a very limited subset of services --- i.e. remote procedure calls. This is clearly very very limited. It doesn't help me if I'm an ISP and I want to offer connectivity and hosting; it doesn't help me if I want to offer a database or storage service with localised caching (which all decent DBs do); it doesn't help me if I want to sell you a system which adds a search engine to your site; and so on and so on. Sure, some of these examples could work via remote procedure calls, but that's not something that should be enforce on them.

Okay, rant finished. For now.

Had more fun yesterday catching up on all the changes over the last fortnight, including a couple of cool new classes from Andy, which allow more streamlined remote management of applications. Previously these were very simple. Now a command can be overloaded with different types sent to it, and they take I/O streams. For example, it's now very easy to stream local a file into a remote application. Given that these comms are also now SSL-enabled that's really fabulous.

A couple of open source developers mailed asking to do something network related, and it appears they have problems accessing Jtrix services through their firewalls. Perfect! We've worked out a really neat way of proxying Jtrix in a variety of ways. Hope they'll be able to work with us on that. It will certainly scratch their itches. And it will also be able to handle SOAP really nicely using the same system. Which brings us back to where we started.

Have spent some time collating responses from our request for open source developers. Pretty good. My job here is to make sure all those people get the help they need. But of 10 or more interested people, I only know of one person who's downloaded the demo. And I don't know what she thought of it. It's odd that people will mail offering help without knowing what they're getting involved with. Nice, of course, but a bit perplexing.

While odd bugs are sorted out in the main Jtrix CVS branch I spent some time updating The Jtrix Dictionary (PDF, 230K) which is a simple desk-side reference to help jargon-busting.

Admittedly, some of the problems I found yesterday turned out to be of my own making, but at least it points up some gaps in how to trace errors better. This one will run and run.

Frustrating day trying to run our various demos with the latest CVS version of the project. This is why integration testing is important! All the parts worked individually, but it doesn't seem to hang together at all. I guess a lot of things changed in the fortnight I was away, and it (almost) suggests that frequent releases is no bad thing, since all these issues would have been resolved in the same time.

Am wondering what kind of interest/activity ratio you can expect on an open source project. We're getting a lot of interest from people who want to help, but they always seem distracted before they even get off the ground. Not very inspiring :-(

Back from holiday! Development over the last fortnight seems to have been good. Our issue tracking system also e-mails, so it kept me informed, along with secure IMAP for mail. This means Day One back at work was very smooth.

In my absence, the team's been working on tidying up lots of things. Beatrix apps run 30% quicker and generally more reliable as a result. Where an application fell over after 2hrs of heavy usage (admittedly an extreme example) it now lasts for days.

This also means some front-end material has changed, so I need to catch up on the documentation changes and then we can release this.

A quiet day yesterday, just lots of heads-down work. Jtrix 0.13 is in test, so we're being careful that it works as expected across the major platforms, and that all the detailed instructions are correct. At the same time some work is on-going to improve efficiency within the hosting service, but that's independent of 0.13. Ah, the wonders of CVS. Hopefully we get 0.13 out early next week.

Re-assuring to see issues being worked out (if not concluded :-) between splork and uleonhardt. Certainly splork has hit on some key issues which may be addressed by Jtrix. Or they may not. At least we are insistent that payment mechanisms should be flexible.

Delighted to get our forum and FAQ on jGuru. Someone posted a question about 2 minutes after it went up. Must be that irresistable "first post" desire.

And O'Reilly's ONJava.com has accepted an article on the platform. Hum, just got write it now...

My article about Jtrix on Advogato seems to have raised its profile, but only a few public responses. I cannot fathom this. Perhaps open source people are so anti-spam they're worried their own words might be considered spam :-) But feedback seems to be pretty good.

My colleague uleonhardt, who is a major developer on the team, tried to reply to a note from splork but doesn't seem to have enough Advogato trust. Ho hum.

We may have our first open source developer, via Freshmeat. We have a few easy tasks in hand, and would appreciate the help, so I've let him know the options and one of our guys is going to get him off the ground.

On the development side we're in the process of resolving a nasty bug related to failover. But since this was working once, and the bug was introduced recently, this shouldn't be intractable.

Couple of (good) interviews on Friday for our site developer position, and have made one offer. It's a sad sign of the times that the candidates so far have been very strong without having to go through an agency.

I also wrote a very short intro to Jtrix at the suggestion of Nelson Minar. It's only three pages. My very first attempt (one year ago) was 40 pages! (It got trashed shortly after, but still...)

Meanwhile our fix for the Windows problem was ready to go by the end of Friday, but it left no time for testing. So we'll do that today and then release when we're happy with it. Would like to get yet another release out before the end of the week to take us to 0.13 because that includes some key visible changes and I'm away for a fortnight after that.

Our push to the open source community continues, and I've put up a list of to do items. We'll do these if need be, but hopefully it will give potential contributors a starting point. We all think Jtrix is great (well, we would, wouldn't we) and direct feedback is very positive, but unless others think it's good enough to spend their time on it sadly reduces our incentive to the open source community.

Several people have suggested some kind of cool app to get people interested, and various options have been floated. At the same time there's lots of groundwork to be covered (see to do list above). And so there's a question of balance: do you spend time on the cool app and away from the basics, with the hope of increasing participation (and hence development) at a later date? Or do you just go off and do the groundwork while everyone else says "Oh Jtrix could change the world, but unless its logo spins round I'm not interested"?

Meanwhile, in a double-edged move, someone mailed to one of our public forums and said they couldn't get it running on Windows. Great! People are trying it out! Oh no! They can't get it running. We'll have a look and get back to them as soon as possible.