The Failed Promise of History-Only CAD
Evan Yares is a guy I have often quoted over the last several years. He’s now the General Manager of Nanosoft America, but he found time to finish up a series of articles he started to write for 3D CAD World entitled The Failed Promise of Parametric CAD. I take issue with a little bit of what he said but the overall message is something you need to pay attention to. The only problem with what he has to say is that the title of the piece, The Failed Promise of Parametric CAD implies that a system like Solid Edge is something other than parametric. It’s possible for a non-history based CAD system to be parametric. I would have called the articles The Failed Promise of History-Only CAD. You’ll get a lot more perspective on the whole issue if you go back to the beginning and read it all. It’s well worth the read, even beyond the topic at hand. Evan puts a lot of stuff in historical perspective, and digs in behind the hype in most cases.
This is not a weakness of a single CAD brand, but of the entire concept of history-based feature modelers.
Evan Yares
Evan is talking about a method called Resilient Modeling System which Dick Gebhard has come up with. I saw his presentation at Solid Edge University last June, and wrote a review of the process on the Solid Edge Community Blog.
The real weakness of history-only systems is the parent-child relationships in feature-based CAD models are not very stable through changes. It is very easy to create changes that break the model’s “design intent”. This is not a weakness of a single CAD brand, but of the entire concept of history-based feature modelers. And it’s not news. If you’ve used systems like Pro/E, SolidWorks, and Solid Edge before Synchronous Technology, you’re undoubtedly familiar with failed features after changes.
Mr. Gebhard’s method is rather detailed, but to boil it down to its simplest form as implemented in Solid Edge, you create the basic prismatic form of a part as a history-free body, and then add details (such things as fillets, chamfers, extruded text) as ordered (history-based ) features.
The message here is not that history is a bad method, but rather that history should only be used for the types of things for which it is well suited. The best method for creating models that react well to changes is to create as many features as you can without parent/child relationships. This would be known as a Synchronous model in Solid Edge, and the Synchronous geometry is shown as gray in the image to the right.
Then add the detail features, usually fillets, as Ordered features, shown in red. If history’s deep dark secret is that it doesn’t handle parent/child relations very well, direct editing’s secret is that it can eliminate topology during changes just fine, but it cannot add topology during a change. This is essentially why we need both forms in a truly robust modeling system.
Any system that is not built with the capacity to create history-free bodies has already been obsolete for about 3 years. And yes, I’m looking at you SolidWorks. The “Move Face” feature does not excuse these parent/child limitations because it is after all, just another history-based feature that gets rebuilt every time the FeatureManager rebuilds, and is thus susceptible to the same failures.
This realization along with the fact that with Synchronous Technology, Siemens PLM has the best combination of tools, is what led me to switch camps. Regardless of what other flaws exist in companies or software, the fact is that history-only CAD is dead technology. The future for full-powered CAD is the combination of Synchronous and history.
SLDPRT and SLDASM files are actually ZIP files. You can’t tell all that much about them just looking in a binary editor… But, if you look using an unzip program such as 7ZIP, you can see everything there.
I haven’t looked through a large number of files, but those I have looked at had no DLLs or other obvious executables. Some embedded files contain clear-text filenames and paths. It appeared that some embedded files containing textual information that might be useful outside of SolidWorks are encrypted, or at least encoded to be hard to read.
Overall, though, what I did see seemed reasonable. I found nothing I’d call spyware.
I have noticed that SW files seem to contain dll software. This easy way of arbitrary extension is an ideal vector for virus and spy software. Great, send a model and infect an entire engineering department. Of course SE is way too smart to allow anything like that. The large size of SE files suggests the possibility of similar problems. Large files are bad in themselves, but may harbor something vastly worse. Of course everything will be better on the cloud.
Excel files have embedded macros. PDF files have embedded JavaScript.
Document based exploits have been around a long time.
I’ve heard people talking about SolidWare file bloat for a long time. I’m curious what all is stored in those files. Ditto for Solid Edge files.
Open SolidWorks files in NotePad, WordPad, or Word, interesting what you’ll see.
Open the file in a byte editor and all is revealed. I just hate the spyware data of all history of all sub directories. Yet SW cannot find files. The best way for SW to fix their files is to include no code, and no crap. The files will be 1200x smaller and easier to share and store.
Data files should have my data, and specific operation data for the programs that will use it. I do not want scripts, dlls, web interactions, or viruses. Microsoft and Adobe are leaders in this kind of bad idea. Maybe not so bad, as their ethics are indisputable.
I am slightly motivated to write a useful virus and add it to my SW files. Such a useful virus might call home and report where it is now. Of course it will be so useful as to add itself to other SW files so that no user interaction is required.
I used the term “parametric CAD” as a shorthand for “parametric feature based solid modeling” — the common way these types of systems have been described for about 25 years.
I knew within a day of posting the first article in the series that I probably shouldn’t have used the shorthand. Because the failed promise is not related to parametrics alone, but rather to parametric features.
(A solid model incorporating parametric dimensions is not the same thing as a solid model comprised of parametric features.)
Oh well… in any event, the “failed promise” (i.e., poor model reusability) is not limited to “history-only” CAD systems. Explicit, direct, or even hybrid modelers can all stumble rather dramatically when trying to make what appear to be simple edits on models that don’t seem like they’d be that bad. (I bet I can easily find a more than a few models on GrabCAD that you won’t be able to edit with Solid Edge or anything else without resorting to hack and whack.)
If you are looking for an easy way to send large cad models then check out Binfer. The link is http://www.binfer.com.
>… already been obsolete for about 3 years.
…and unfortunately that doesn’t appear to stop the same people recreating the same thing anew to be obsolete even before they start. No, wait, I remember, its the cloud that opens up all sorts of CAD possibilities isn’t it? Yes Belmont Tech looking at you….
Neil,
I know they were working on a cloud thing, but have they said it’s a history-based tool?
Well its based on Parasolid we know. I doubt SE will give them Synch since DS couldn’t have it. It wouldn’t be smart of Siemens to do that. Even if they cant market SE I don’t think they are so stupid to give it away, and how long have SE been working on their implementation probably 5 years already… Its hard to imagine Belmont can front up with both cloud and non history from the outset even if they wanted to. From their limited press and hiring indications they seem to be fixated on the cloud as being the new development branch for the industry rather than as a selective augmentation of it. Remember some key people are conspicuous cloud apostolates and they have picked up venture capitalists with a bent for that purpose. If you listened to that recent JH video then you know their goal is to provide text messaging and skype for freshman engineers foremost and that also incidentally does CAD. I’m joking but that’s their arse about face reasoning/approach. I am picking cloud remains their focus rightly or wrongly. I don’t think they are in touch with reality, understanding the times or on the ball at all. They already need to produce a whole new UI ,documentation, sales network, web presence and cloud tools… They are going to need to license FEA CAM from somewhere as well. Autodesk might end up being the only landlord in town there. However when are they going to have something useful enough to sell against whats already available and mature for pro use? Even if they had Autodesk’s resources and established code to work from it could be 7 years in the making. What will they have to offer that’s unique at that time? You cant really expect something limited to fillets no matter how elastically and remotely they are done is going to cut the mustard or generate enthusiasm. Could you drip feed a fledgling toolset into the market and expect to pick up a credible and sustaining user base from ex SW, Inventor, SE users I wonder. If SW users don’t go for the cloud now why would they be interested in Belmont’s offering of SWII in the cloud then and being so new in every aspect its going to be massively bugged? Many unanswered questions, but common sense tells you some aspects of their mission are fundamentally flawed or misplaced. It cant all ride on the repute of a small group of ‘has beens’ recreating glory days in the new remote age surely. In 7 years how will the landscape look anyway? That’s a whole lot of winks of a young girls eye. My bet is that the West is in the midst of a prolonged and deep reset. I don’t see any growth for decades. I’m really not sure diverse collaboration and fast product cycles will be the vehicle for industry in the post collapse phase. If SW users want an alternative to reposition themselves for a brave new world I think SE offers the most likely route today what ever else may happen along. Now if only the UI was a little better and it did more surfacing. Belmont should have put their money and talents into perfecting their parent technology. That would have been a good combination. Decent marketing and an intuitive UI for SE…. 😉
Want to know how much I know about what Belmont is doing?
Not much.
OK… I’m pretty sure that their product will feature a scalable architecture supporting cloud deployment. And Siemens has already announced that Belmont has licensed Parasolid and D-Cubed.
But that says almost nothing. If you start with a cloud architecture, you can deliver it on a workstation, if you want. (You can’t do it the other way around.) And if you start with Parasolid and D-Cubed, that doesn’t prevent you from incorporating other geometric modeling technology. Including both direct and history based modeling. Or maybe something even better.
Truth is, I actually talk to the guys at Belmont every once in a while. We don’t talk, in specifics, about what they’re doing. But I can tell you that there are no has-beens there. And, no matter what they may ultimately deliver, it’s likely to raise the bar, and help drive innovation throughout the CAD industry. Even in Huntsville.
Nonsense!
Solidworks generates parent child relationships that are unnecessary for the definition of geometry. A part that consists of several merging features usually does not care about the order of the merges. Solidworks is too stupid to allow reordering of the features to help clarify the design intent. The deletion or failure of one feature can vanish dozens of surface operations.
I am so pissed at SW today. A complex surfaced part refused to rebuild after no changes. A split line, knit thicken,fillet would no longer subtract from the base body. I remodeled the feature with projected sketch on suface, 3D sketch for corners, GW least tension blend curves, 3D sketch with copious convert entity, sketch for groove profile, a couple of cut sweeps, mirror features, and a few SW crashes.
I fear that the root of the problem is in the kernel handling of trimmed surfaces. I try to use features that produce only untrimmed surfaces. The best surfaces are those that are defined with 4 edges each of which is a beautiful spline or conic curve. Rather like Tsplines with a very small number of patches. I hope SE does this better.
Rick
I have just finished the prototype phase of a new handheld product. Lots of surfacing. I have not run into all of the problems that I did using SW doing similar designs. The surfacing for a first release is much more stable and seems to have a high precision level. Meaning no jagged surface intersections and much better performance and stability. I have NOT crashed once utilizing the surfacing tools. I have used several betas and am still running the last beta (Rev 1 release) and I haven’t updated any service pack yet.
I don’t think its the kernel due to NX surfacing uses same. Its implementation.
For a first release the surfacing pack is pretty amazing. And it can only get better.
The changing of upstream features etc which cause so much grief in SW seem to be much less harrowing experience in SE.
Dylan,
It isn’t common sense for people who have fallen victim to marketing hype. Some systems promote a “fast and loose” approach, where you’re allowed to do just about anything, and of course there are no consequences for bad practice. Some even codify this in their official training documentation.
But even good practice can’t rescue a system built on a flawed concept. We lived with the flawed concept for the last 20+ years because we didn’t know any better. But now there is newer better stuff that takes better advantage of what we’ve learned, and of course the improved new hardware.
I participated in the VAR training stuff for years, and yes, I believe it is mostly overpriced for what you get, although it varies from source to source. Some instructors are truly cringe-inducing, while other have a lot of valuable insights beyond the official training materials. As Dave says, some sources of information have a better bang for the buck.
It’s worse than you think Devon. University of Alabama Huntsville, you know the one right next to a bunch of world class cutting edge technology companies, still does not teach ST. I can’t fathom a university professor standing in front of a class who is supposedly being prepared to sally forth into the work place prepared not teaching this.
We had some students at a Huntsville user group meeting that were in awe of ST that they had never been shown at UAH. I guess I am closer to Dylan than these students though because in spite of either a very lazy or a very incompetent or both professor I would have been curious enough to have looked into ST on my own anyway.
And dittos on the cash cow of VAR’s and training. I looked at an excellent Saratech video on CAMWorks for SE and their special offer at the end was one on one mentoring for CW4SE at something like $192.00 per hour. I gave a VAR money years ago for training. It was $2,400.00 for four days because there were enough users there to “reduce the cost”. No videos, you could not record anything and they threw so much at you in four days that I determined I would never waste my money with these VAR rip offs again.
I recently purchased for around $50.00 a PDF with embedded videos on CAMWorks for SolidWorks to learn about CW4SE since they are so similar. Step by step and 80 minutes of videos showing you exactly how things were done and you can read and view time and time again. I like this method of learning.
Matt-A well written article, thanks.
RE:”I learned to model in this manner through trial and error, and never received any official training…Is this approach to modeling not taught to students/trainees?”
Dylan brings up a good point. In my experiences, the majority of SolidWorks users were self taught. One reason being the very high cost of SW Reseller Training. I taught some of these classes & the users were charged as much as $500-$1,000 per day, ridiculous IMO. Of course they were told to come back every year for more training.
So when I would consult, especially to mid size to large companies, I would see many different methods & styles of working using SolidWorks. This lead to an enormous amount of wasted time as users used the incorrect methods or argued about it over & over.
A new method of training CAD users could be a game changer, IMO.
Cheers, Devon
Every time I read about this, I think that Dick Gebhard’s Resilient Modeling System could be boiled down to two words: common sense! I learned to model in this manner through trial and error, and never received any official training…Is this approach to modeling not taught to students/trainees?