Raj

Instead, we looked at the rules and designed a home grown solution that fit the requirements. the rules were rather simple, but the tricky part was allowing multiple versions of rules without having to load 500 rulesets and dynamically looking up the rule service. I actually spoke to mark and michael about this. don’t know if they plan to support this type of functionality or not.

Sorry that I have not responded to my blinking lights and computer whirring noises recently but I cruising the net with an older G4 that I normally use only for testing, research, etc. Now – on to comments about comments.

Peter – I tried to answer most of your questions but apparently you still have some. If you know Don Tallo at Fair Isaac, you might drop him a line or two about your problems with Blaze Advisor. I’ve known Don for a really long time and he’s always been (to my knowledge) straight up about the product’s shortcomings and benefits. Also, he would be the person that I would ask if I had a question since most of the Professional Services folks are NOT really AI guys, they’re Java programmers who have learned how to “get around” in Blaze Advisor better than most of the customers. Don is a true “AI Guy” who started the hard way and move up the corporate until he got to the point of being a manager and sacrificing 20+ IQ points or staying technical. Much to the loss of our industry, he chose money over fame and lost his points. But he had plenty to spare and is still the “go to” guy for Blaze Advisor technical questions. Even though he does not surf blogs such as this, he might respond to an email. His well-known company email address is dontallo@fairisaac.com – tell him that I sent you.

Daniel et al – the point of the whole discussion was that Drools and other “free” software is following a well-known trail to the Money Tree. First, have a good product that performs as well, or almost as well, as the commercial products. Then put on the “window dressing” that is needed for a commercial product in the name of “being helpful.” Once all of the bugs are worked out, go commercial, close off the open source stuff, give the inventors and hard workers a pat on the back as you show them to the door and slam it shut behind them. After that, the product becomes a “hidden source” just like all of the others who have gone down that path.

Only Linus Torvalds had the intestinal fortitude to resist the allure of the dollar. He owned it and he wouldn’t sell it. He did license the engine to commercial companies such as Red Hat and Novell. Unfortunately, Mark Proctor, one of the high holy priests of Drools, is NOT a sufficient stockholder in Red Hat to keep them from doing things the “commercial way” once he has finished the product to the commercial level. When it happens, remember me, this page and my prophecy .

The funny thing is that neither Mark Proctor, Michael Neale nor Edson Terilli have commented on the blog.

Daniel: BTW, my name is Yaakov Kohen. Translated into English Yaakov is James or Jacob; either is correct. In German it is still Yakov in pronouncing but usually spelled Jacob. Translated into French it is sometimes Jacque. But for this page, it is Yaakov. I’m looking forward to seeing all of you at the Dallas Rule Group conference, October [Technical] Rules Fest. Thanks for all of the comments,

SDG
Yaakov

Interesting discussion. I agree with Peter that although we have come a long way since the raw expert system shell of the ’80s but there is still a long way to go, and probably a few forks in the road ahead. There is no shortage of customer requirements or ideas for cool new innovations!

In my opinion the major challenge for BRMS is enabling diverse groups of people that interact with application logic to deal with change in a structured and coordinated way. BHAG! It so happens that a rule engine is a great tool to help with that problem. Of course the challenge of integrating the design time and runtime into the tools end-users want to use is constant and evolving. In the time I’ve been at ILOG I’ve seen the focus be on everything from Swing applications, to the Eclipse IDE, Eclipse RCP apps, Portals and portlets, Web Services and SOA, AJAX mashups, MS Office and .NET rich clients. There will doubtless be more as we see the distinction between the desktop and the web get more blurred.

We’ve got some great stuff in the pipeline, both for the design and runtime aspects of the BRMS, so we plan to keep raising the bar for BRMS products. Personally I’m really interested to see how the various products in the market evolve, based on a fairly diverse set of requirements and a complex competitive landscape.

Dan

peter

The multiple rule versions is a common requirement I’ve seen in health insurance, where an application needs to run different version based on the data. For example, if a claim was started on 10/05/07 with rule version 3, but the current version is 5, the application needs to use version 3. As far as I know, Blaze does not support this feature. Likewise, JRules does not support it either. JRulse recommends you look up with ruleset to use, but that is prohibitly slow and not practical. If I have 50 rulesets with 8 versions each, it means the ruleserver has to potentially manage 400 rule engine instances, or dynamically load it each time. For an application that doesn’t have much load, it’s probably ok. For an application with high availability requirements, it isn’t going to be practical.

I’ve tried the rule dependency chart in blaze and I’m not impressed. What blaze provides doesn’t have the granularity I’m thinking of. I want to see the dependency down to the conditions, functions and actions. I asked a FI consultant onsite at the last project I was on and his answer was no. Is that possible? and was the consultant wrong?

I know that said tabet has used filters with knowledgebase systems in the past, so it is needed by some people. I don’t have any studies to show the percentage, but I do know there is a need for it. One area in particular that could benefit from it is data mining. Many people do it manually today using simple business rules or decision trees, but it would be more efficient using kalman or bayesian filters.

In terms of distributed Rete, I’ve seen a need for it in the past on 2 different projects. The reason said and I worked on it is because we saw a need for it. It just wasn’t feasible with JESS or JRules.

On embedding, what I meant is embedding the BRMS in an existing application. By that I mean say I have an internal application for managing the call center. Say I want to embed a the BRMS so that business users can write rules, test them and management deployment. Today, with blaze and JRules, you have to install it as a separate application. This means users have to use 2 or more applications to write, test and deploy the rules. Embedding the rule engine is easy and all of them support it. Embedding the BRMS application isn’t a feature that comes out of the box.

I’ve looked at visual rules in the past, but I wasn’t aware they had a BRMS UI designer. I know pega has a nice UI designer for the PRPC product. Since PRPC is a fully integrated product, the UI designer handles the BRMS and application at the same time. The downside with Pega is the terrible rule engine underneath it. The way Pega is designed, it won’t scale with increasing loads or concurrent users.

In the past, I wrote custom rule editors to make the user experience more friendly. Rather than use multiple tools, the users could write the rule within the same application, test it canned data or real data and manage the deployment. We call it “what if” rule testing. Basically a compliance officer could write a rule, take an existing account and rule it. The application then returns a report with full details and explanation of what went right/wrong.

As much as BRMS has improved over the years, my experience has been that current tooling only solves half the problem. The other half often ends up being quite challenging and difficult. In that area, I think BRMS could improve to make it much easier.

peter

I learned a long time ago that when a sentence has a “but” in it that you can forget everything before the “but” because it’s just fluff to hide the real meaning after the “but”. So, to answer your comment in a logical fashion, have you published your comparisons? If so, would you provide our readers with a link to them? Thanks.

I, too, have been doing some comparisons of various tool vendors but I haven’t made the public except through InfoWorld articles. However, the ability to run multiple versions of a single rule in a single rule engine instance is something that I’m not sure is needed – at least, this is the first time that I’ve heard about it being needed.

Decision Trees normally have rules that support them and usually can be compiled.

Fair Isaac has a dependency chart (not sure about the graph part) for rules in one of their reports.

I’m not sure how many times that I’ve been asked for Bayesian Filters, Kalman Filters, Complete Temporal Logic Support, Simulation Framework and Distributed Rete by business folks. Oh, right. None. (BTW, it’s Rete, not RETE.)

Actually, rulebased systems have been around since the 60’s – I’ll put some early papers by Alan Newell and Herbert Simon on my site later this week – but the commercial applications only began to get cranked up in the 70’s and early 80’s so I guess you were pretty close on that one. But to say that they will continue for the next 30 to 40 years? I’m not sure why you say that but I hope you’re right. True, everyone predicted that COBOL would die out by 2010 but it’s still going strong. Personally, I look for a marriage of Neural Nets and Rulebased Systems to form a true AI KnowledgeBased System – something that Doug Lenat has been working at since about 1984 – or was it ‘89? Regardless, he has predicted that Cyc won’t be finished until 2025.

Regarding your point about a BRMS with a UI designer – that’s what Visual Rules (that does NOT use a Rete engine) has done. They have used Eclipse and allow anyone to redesign the interface to meet their own needs. I’m sure that others will follow suit before long.

Finally, embedded BRMS – Oracle has embedded Jess. I personally know of several companies that have embedded either JRules or Blaze Advisor. I’m sure that Mark Proctor could provide a name or two of companies who have embedded Drools in their applications.

So! As always, enjoyed your comments. Keep those cards and letters coming. Proverbs 27:17 says that, “As iron sharpens iron, so one man sharpens another.” Monologues are for comedians, dialogues are for learning.

SDG
Yaakov