From Russia with CAD
All articles     

13 Sep 2010

Discussing LEDAS’ heavy technological karma with David Levin

CADOVOD is one of the most noticeable and bright bloggers writing about CAD in Russian (Cadovod in Russian sounds like a person who is cultivating CAD - similar to cultivating vegetables or something…). Recently, he published an interview with David Levin, the interview which might be useful for those English-readers who are interested in history, current positioning, and development trends of the LEDAS Company.

I continue my talks about CAD with interesting persons. Today’s guest of my blog is tri-personal: he is the head of a software company, the editor-in chief of the only CAD-portal in the Russian Federation and certainly the most famous Russian CAD-blogger (he also supports a small English blog. However, our talk with David Levin will focus only on his company - LEDAS as, in my opinion, LEDAS is not well known to ordinary users in Russia. And it’s definitely an unorthodox company. Our conversation was so long that unfortunately I’ve had to squeeze its content to fit the blog format... Nevertheless, you’ll find out how hard it is to stop able mathematicians, why Top Systems does not need any solver, whether and when LEDAS is going to release its own CAD-system, and lots of other interesting things.

David Levin in his office

In my opinion, LEDAS is little known to general public in Russia. For instance, I’ve never read anywhere about LEDAS background. So how did it start? What did you do before founding the company?
We have an extensive background and we have come a long way: from an artificial intelligence laboratory through participation in a real and a very well funded Soviet project for developing a new generation software-and-hardware system to organization of our own research institutes (a unit of Academia and an industry facility).

Our research projects focused on automated natural language understanding, parallel computing and new software architecture. The work met the highest standards, which translated into serious publications in foreign scientific journals, participation in conferences and western scholars frequently visiting us.

Our special achievement was implementation of an original system of constraint programming. Effectively, 20 years ago we already had a working algebraic solver that performed calculations with sets of equations and inequalities with any type of parameters – geometric, mechanical, etc. Although such a universal approach certainly affected the productivity level, it is difficult to overestimate the role of competence and experience accumulated by our team.

Still, what compelled you to form your own company?
I would say, objective preconditions were already there around 1989. But companies are founded by real people, and real me needed ten more years to feel, if stated pretentiously, an insurmountable striving for independence, creative latitude and even, I dare say, my personal responsibility.

I think such feelings are decisive for many company founders. My desire and willingness to form a company (and generally, my business “sparks” and education) were heavily influenced by nearly 3-year professional and personal contacts with top executives of an American company Exigen and their Moscow partners, who at that time made serious investments in our projects.

Understand. And how did your co-operation with Dassault start? It sounds fantastically that a small Siberian company has become such an important link in development of world-class CAD-solutions.
Perhaps as any company of such level, DS closely and continuously monitors international publications on relevant science-and-technology issues. Through our research publications, they found us as experts in constraint programming. For a long time we mostly were commissioned to work under small-scale research contracts and although they thought highly of us as smart guys, at that stage our cooperation had good chances not to survive.

The breaking point was reached when – in the mode of friendly information sharing with our respected partners – we showed our DS colleagues the first version of our own geometric solver.

Which year was that? Wasn’t the mode of friendly information sharing a ruse of war from your side :)?
In 2001. Yes, it was :). They found me straight away although I was on vacation, suggested I should immediately meet DS top executives and then offered LEDAS quite a large and important commercial contract. Since then the volume of contracts has increased considerably, while the level of responsibility has not diminished.
Could Dassault simply acquire LEDAS? Doesn’t such high level of cooperation with DS create certain obstacles for work with other companies?
We discussed and continue discussing regularly development of our relations. Documents regulating relations between LEDAS and DS do not put any formal obstacles for our work with third parties. So far we have been able to easily overcome any informal concerns of some our clients by means of security supplement well-known to lawyers as de bene esse.
OK, the situation with Dassault is clear. Don’t you find it strange that among Russian CAD-vendors only ADEM uses your solver? Is it wise for ASCON and Top Systems to “fragment forces”, each writing its own solver? It could have been a useful co-operation with labor specialization :).
There is nothing strange in ASCON using its solver. There are at least three simple reasons – first, their key product KOMPAS could not wait ten years for LEDAS releasing LGS, and to change a solver in an already top-selling CAD one needs serous reasons and considerable resources. Second, if a company has its own able mathematicians, it’s difficult to stop them. Third, I’ve already talked about striving for independence and it is not the quality typical just for Mr. Levin :).
What about Top Systems?
Top Systems also could not wait for us. And they certainly have talented developers. And independence is undoubtedly not unfamiliar to them. There is, however, a fourth reason – Top System simply does not need a solver: they consistently advocate a history-based approach even where nearly all leading vendors abandoned it in favor of constraint-based modeling.
Are you suggesting that Top Systems are moving towards the wrong direction?
From my end it would be kind of shallow to think that a particular approach has conclusively taken over. But discussing comparative advantages and shortcomings of both approaches is clearly beyond the scope of this interview. If interested, you can read a key article of Dmitry Ushakov on "Variational Direct Modeling: How to Keep Design Intent in History-Free CAD" or a series of articles by Paul Hamilton on "Editing 3D Geometry", where these questions are discussed in detail.
So what’s the best: buy a solver or develop one’s own solver? And if suddenly all decide to have their own solvers, what will you and D-Cubed do?
So far I was only trying to give a reasonable answer to your questions about ASCON and Top Systems. But I have a more general and a more radical standpoint with regard to the issue.

Answering the question “which solver is sufficient for a particular CAD” in the “your own – someone else’s” dilemma, I believe it is necessary to take into account this important factor: what is an expected development capacity for a particular CAD, as well as for dependent components and applications? I think that a globally competitive CAD-vendor must be prepared for new (perhaps, radical) market requirements and by and large must be able to foresee such requirements. In my opinion, this puts much more serious development requirements (architecture, algorithms, level of implementation… and even concepts) than in local assessment of company’s own solver.

To make a simple metaphor: a solver used by a globally competitive CAD-vendor must have the features of products capable to expand on the market beyond its own CAD. Not to expand, but have the features of an expandable product! To attain and maintain such features, a company needs considerable resources. LEDAS experience shows that for your solver to reach the level of nonrandom sale to the global leaders, you need to invest around hundred (highly) qualified man-years.

Still do you think it is normal that all or majority of global CAD-leaders and simply big vendors put efforts to develop their own solvers?
Here I’d like to make an even more radical statement, referring to your recent interview with well-regarded by me Vladimir Zakharov, director of software development at ASCON. By the way, what’s funny in that interview talking about "leaders and solvers" interlocutors pretend that Autodesk does not exist. A well-known method to overcome a competitor? I think results can be opposite to the intention.

So, in this interview a rather firm statement was made that all Big Four companies (or the Big Three as Mr. Zakharov believes) have their own solvers. It’s far from being correct. I cannot give any detailed comments to this statement as I’m limited by tight NDA-obligations but those who want can trust me on this.

I am not even prepared to think that Siemens PLM Software has its own solver, or, more precisely, that UGS had its own solver. As known, UGS first bought a solver and then made this solver its own by acquiring the solver’s producer. There is nothing wrong in such solver production – do it if you have enough money:).

As for (very) large companies that are not members of the Big Four, the list of D-Cubed clients shows that hardly any well-established, reputable company produces its own solver. It is also confirmed by the expanding list of LEDAS clients, of which not everybody, like CD-Adapco, allows to announce licensing our solver.

Now a reverse question: have you ever wanted to grow to your own CAD? You have the solver, you can buy the kernel, and the rest is paperwork:). Or may be this idea is already embedded in your future plans? Does bCAD - a Novosibirsk product - have any relation to LEDAS?
According to the LEDAS experts, even to license ready components of third-party developers (Teigha/RADF, HOOPS/Redsdk, ACIS/Parasolid, DCM/LGS), a company that intends to create its own CAD must invest around US$ 5-10 million in R&D, marketing and sales channels. We will not be able to pull off such a project without outside investors.

On the other hand I suspect that today it is practically impossible to spin off a new CAD on the market. It is illustrated by an example of SpaceClaim: even with solid investments, particularly, in marketing, and Mike Pain‘s charisma and connections, even with a wonderful technological idea and its implementation – looks like their product is still not breaking even. To a considerable extent, AutoCAD clones also demonstrate difficulties of such development.

Sometimes LEDAS suffers from a short-run loss of common sense and then the thoughts come: wouldn’t it be awesome to build our own CAD, with detailed knowledge how CAD work inside:),and obtaining yet another dimension of independence :).

Regarding bCAD: ProPro, who made the product, is our neighbor. They have always been just several dozen meters away from us and we even share close academic roots. Throughout the life history of both companies we discussed the prospects of cooperation and even worked on some miniature projects together. It so happens that one of the ProPro founders – its former Director Vladimir Malukh, and two other young developers – former members of ProPro staff - now work for LEDAS.

Well, why didn’t you build your own CAD based on LEDAS and ProPro experience!
Sure, we had this screw loose temporarily …:)

Let me better answer your question about own solver more broadly.

Voluntarily we will not abandon our position of the only real competitor of Siemens PLM Software (see "LEDAS Responds to D-Cubed" at upFront.eZine of Ralph Grabowski) on the global market of geometric solvers. At the same time, we do not intend to always be only a supplier of components for developers and unique outsourcing services. Not least because even a very cool business for production of high-tech components is not so profitable as selling sufficiently mainstream end-user solutions.

So already in 2008 we announced developing software products for end-users under the general brand of Driving Dimensions. These are not ready CAD but plug-ins to popular systems such as SketchUp and Rhino. Our plans include other target platforms. With Driving Dimensions plug-ins, users of these systems can employ easy-to-master parametric design tools. Putting Driving Dimensions to the market, we also want to popularize our variational direct modeling technology (which combines direct editing of 3D geometry and its parameterization by means of constraints), see some LEDAS white papers. However, this is secondary in comparison with the main business objective: to become a supplier of end-user solutions. We have already developed reseller networks in 16 countries throughout the world, and in some of them the interests to our end-user products is so high that right away several companies decided to cooperate with us in sales. This way we supplement our components for CAD developers with a plug-in portfolio for end users. Further on we’ll see how successful our business is; perhaps one day we’ll feel strong enough to enter the market with our own CAD-system, but at the moment there are no such plans.

Still, paradoxical development of events is possible. Today we have already somewhat achieved our secondary objective – popularize our technology by means of end-user solutions. Recently we have been approached by serious clients who saw the results of Driving Dimensions and who are interested precisely in this technology. And if the client makes an offer you can hardly reject? This way we may put our dream about selling mainstream end-user products on the back burner and re-enter our heavy technological karma …:)

This karma is also highly commendable… Thank you for your time and let me wish you every success in your projects!

See also:

Permanent link ::

collision detection
23rd of June, 2024


All news


All articles

© 2004-2024 LEDAS Ltd. All rights reserved.