Pure was an amazing product. For C++ development, we relied on it for my team at SAIC, and when I moved to Angel Studios (we developed games and content for Nintendo, Sega, Disney) I immediately asked the owners to purchase a company license.
I saw Reed Hastings being interviewed on YouTube last night - fascinating how he has so many personal friendships with CEOs of competing companies.
Purify could find memory errors in just a few minutes that would take engineers days or weeks to track down. The dynamic instrumentation capability on existing binaries was huge. The first time I saw it in use was at Sybase. Running a demo version for an hour found enough problems that we instantly sprang for the license.
I used it at Sybase when I interned there in 2009. I've been fascinated by good developer tools ever since. It was such a joy to run, understand the reports and make fixes, and repeat.
Purify was awesome back then, before Linux and the open source tools.
(I think Purify might've been the first phoning-home tool I saw, finding a way to email Pure if you exceeded your simultaneous licenses, but I might've misunderstood at the time. Around that time, it was normal for big-ticket Unix workstation software to have network license managers, which were tied with stale NFS mounts at our site as the most common cause of a fleet of expensive engineers on expensive workstations with expensive software all suddenly unable to work. :) )
I worked at Rational Software when we acquired Pure software. We sold tons of Purify+. The legend of course was the Hastings took the $ he made from selling Purify to Rational to start Netflix. So as a Rational employee, I had a contribution to Netflix's existence... Or so I like to tell myself ;).
Pure products were amazing. I always found the way Quantify represented profiling to be incredibly helpful back in the ancient days. I know there’s more advanced visualizations but I always found the logarithmically scaled directed edges in quantify to speak to me better than flame charts and other related techniques, I think because it represented the full execution graph rather than individual transactional paths. Anyway huge respect for Reed as an engineer and have always been impressed with both Netflix and the quality of the performance, debugging, and other efforts that have come out from them. He clearly values excellent engineering.
I kagi’ed around and I can’t, sorry. The software is 30 some odd years old and was acquired by IBM and sublimated as all ibm acquisitions do. But my memory of it could well be approximated by using dot to create nodes for execution points with edges to calls that are rendered with a weight proportional to the log of the execution time. The nice thing is it allowed for recursive calls to be rendered as cycles, and there are a lot of non recursive cycles I found. It also did a reasonably good job of clustering so highly popular calls fanned into a sub graph but with each graph calling in clustered independently.
Is there anything to support that history? Documents from that 1997? Considering the company was founded by just trying to come up with random ideas I doubt they had too much foresight.
They were a tech company of their time: they had an excellent web site, search (and recommendation IIRC) etc, especially compared to most of their contemporaries.
Nowadays I wouldn't consider anyone in the DVD-renting business to be a tech company. I would still consider Netflix a tech business under their current business model.
Note that Netflix is relatively small and not really in control of their fate; they were presumably included in the "FAANG" acronym to make it fun. Now that two of those companies have changed their name it doesn't even fit (A, A, A, M and...N? Not the other M?)
Netflix was included in FAANG because it was a high growth tech stock, growing faster than all the others when it was included. It was the fastest growing stock on the market multiple times in the 2010s and was the fastest growing stock for the entire decade.
"Nowadays I wouldn't consider anyone in the DVD-renting business to be a tech company."
Netflix still rents DVDs.[1]
I actually much prefer their DVD service, because their selection of movies on DVD is infinitely superior to their streaming service, whose selection is absolutely atrocious.
Netflix's streaming service might have world-class delivery, but what they deliver is crap.
Recommendation accuracy was much more important for DVD rental, because of the wider selection of content and the slower feedback loop. So a good recommendation algorithm was a core competency for the business.
And when seeing his internal recommendation algorithm team hitting the wall in recommendation accuracy, Hastings boldly created a 1M$ "Netflix Prize" competition to create a better recommendation algorithm [0]. This boosted research efforts in the field and earned him and Netflix respect in ML academic circles.
Netflix was originally intended as a streaming company. The DVD rentals were a strategy to solve some non-technical problems with the business while the infrastructure for streaming caught up.
That's fascinating and an incredible example of long term planning. Do you have a source for this?
edit: A very cursory check of the Wikipedia page and https://www.vox.com/2017/9/13/16288364/streampunks-book-exce... suggest that it probably isn't true that Netflix was "originally intended as a streaming company" considering at it's founding online streaming was still far from feasible, but I'm open to evidence to the contrary.
My source is that I worked at Netflix in the early days, it was always part of the strategy. We were discussing streaming internally almost a decade before they actually delivered it.
While streaming technology was early and teetering on the edge of viability, the real challenge was maneuvering the studios to license content that they didn't want to license because they wanted to retain absolute control. They were very averse to streaming for multiple reasons. The gambit of using DVDs in a series of Hobson's choices for the studios almost worked but ultimately failed at the late stages for an important reason that was difficult to bridge.
From Netflix's IPO prospectus, dated May 22, 2002:
> We currently provide titles to our subscribers on DVD only. However, we continue to monitor additional delivery technologies and, when appropriate, believe that we are well-positioned to offer digital distribution and additional delivery options to our subscribers.
And further down, a bullet point in their growth strategy section:
> Implementing Digital Delivery. We continuously monitor the development of additional digital distribution technologies. Historically, new technologies, including the VCR and more recently the DVD player, have led to the creation of additional distribution channels for filmed entertainment. We intend to utilize our strong relationships with the studios to obtain rights to acquire and deliver filmed entertainment through emerging digital distribution platforms as they become economically, commercially and technologically viable for those subscribers who prefer digital distribution.
And later, regarding competition:
> We believe that our strategy of developing a large and growing subscriber base and our ability to personalize our library to each subscriber by leveraging our extensive database of user preferences positions us favorably to provide digital distribution of filmed entertainment as that market develops.
Streaming on the internet was already exploding by 1997; relatively speaking, of course. I distinctly remember first watching South Park--myself and several other students, working in the computer lab technical assistance office for work-study, huddled around a Mac. This likely would have been circa 1998-1999, if not late 1997 when it debuted. (In fact, I don't think I ever watched South Park on cable television until many years later, while staying at hotels.) Around that time I also remember a major, well-funded streaming website that was supposed to revolutionize Indie film distribution. Anybody working in computing or the film industry knew where distribution was heading--the only question was when all the pieces would come together to make it viable, broadband adoption being the most important. (Not sure if a majority of American households were yet "online", but if not that was already a rapidly approaching inevitability.) There were many ventures which dove in too early.
While I wasn't paying attention to (or even knew much about) Netflix before it's IPO, certainly by then it was well recognized--and recognized as brilliant--that Netflix's strategy was to build relationships with studios and rights holders on the one hand, and customers on the other, precisely to position themselves to dominate digital distribution when the time came. Online DVD rentals was the ramp, permitting them to build momentum so they'd have a head start ahead of anybody else in the digital distribution space. And it worked spectacularly; it took nearly a decade, if not more, for anybody to even come close to Netflix's lead. (Wikipedia says Netflix began streaming in 2007, same year as Hulu, but Netflix was still dominating until only a few years ago.)
are you a young un? dont you guys remember real audio? i remember streaming the NBA playoffs over dialup all the way in Australia. Video streaming was already a thing, it just didn't have a killer library.
> It was an online website right around the dot com bubble.
Much sooner than that.
Amazon started in 1995. Not a lot of ecommerce websites existed then. One the earlier ones is Powell's Books which started selling in 1993 via telnet, in 1994 via web.
Tech company did and still does to an extent refer to "online" first, vs brick and mortar. In the context of the dot com boom, they were definitely tech
it absolutely was a tech company. the 'Cinematch' recommender , replaced eventually, was one of the differences brought-up in the netflix/blockbuster lawsuits.
I have a copy of the Pure Software Engineering Handbook which was created for internal use only. It's amazingly comprehensive and well-organized. I've never seen any kind of onboarding that comes close.
Having met him before, his clarity of thought just comes across in this paper as well. Some people are just wired with that talent that makes me so jealous. Reed is one of those people with high EQ and IQ (with the ability of handle his inner Elon diplomatically)
You're citing an example of high EQ (unless you think he really means it, which is doubtful). It won't hurt him in any way to say that and he gets credit with Elon and his fans.
Reed Hastings knows too much about creativity to believe that. I don't condone this kind of fake praise, but it is easy, costless and will make you a "good guy" to a little demographic that can be troublesome if aimed at you. Even if the public all sees through Elon and associating with him becomes anathema, it won't bite him at all.
I'm reminded of some commercials that kept playing on YouTube a few years ago where John Mayer was calling Kanye the most creative genius he knew. I'm sure he didn't mean it, Kanye is not someone you want to endorse now, and no one cares that Mayer said that.
You don't like Elon and expect people you hold in esteem to agree even when they state the opposite. I know this is mind-blowing but there exist smart, talented people who disagree with you. And not just about this.
Also, the claim that praising Elon is a safer play than condemning him is not obviously true. There are plenty of Elon haters out there who could be "troublsome". Like you and the other guy who lost his shit in this thread.
I don't really dislike Elon. I think he doesn't do things from a bad place but possibly because he's not neurotypical he's been sucked into spending his energy pleasing his craziest fans.
We'll have to disagree on Hastings quotes, to me they seem clearly a way to speak nicely about Musk with a wink to those who don't idolize him.
> You're projecting your opinion onto someone who doesn't share it.
Two things:
1. interfacing with politicians (I'm including both executives and senators in this group from personal experience) has taught me that many are exceptionally diligent with the words they use. "Brave" and "creative" are neutral adjectives politically even if they have positive connotations interpersonally. If a politician doesn't say something exclusively, denotationally positive, it's almost always a deliberate hedge. I'm not saying that's the case with Reed, but I'd bet a dollar that it is, because most people who's impressions he'd likely care about (execs, congress, etc) use and read the same language.
2. Unless you spoke with him and heard otherwise, aren't you likewise projecting?
No, I'm not projecting, I'm taking what he said at face value...
He didn't just say "brave" and "creative". He said more than that, including some semi-negative comments about Elon's general behavior, but the thrust of his comments is very positive.
It's absolutely hilarious to see people who hate Elon react this way. You can't compute someone you respect disageeeing with you.
> It's absolutely hilarious to see people who hate Elon react this way. You can't compute someone you respect disageeeing with you.
The fact that you jumped to this conclusion when I said nothing to indicate how I feel about Elon in my only comment on this entire thread says a lot about where you're coming from.
Anyhow, the fact that his comments were:
"[I am] 100% convinced that he is trying to help the world in all of his endeavors"
"He's trying to help the world on that front because he believes in free speech and its power for democracy,"
and
"how he goes about it is not how I would do it."
should give you a pretty clear read about where he lands right now:
1. Reed at least publicly believes Elon is coming from a good place
2. Reed disagrees with the methods Elon's using.
3. "Brave" and "most creative" contextualized by the statements above are given a neutral tone and meaning.
Tldr: Reed basically said "he's coming from a good place but his approaches are bad"
I read you correctly. Congratulations on looking up the full comments (or at least more of them) but you're still reading it wrong because you can't handle the fact that someone smart disagrees with you.
Hastings added that he is “100% convinced” that Musk is “trying to help the world
in all his endeavors,” and added that he was “excited” about Musk’s Twitter
acquisition. Musk “just spent all this money for democracy and society to have a
more open platform — and I am sympathetic to that,” Hastings said.
I may be in an overly cynical mood, but the word "trying" in that phrase changes the color quite a bit, and frankly, the word didn't need to be there. Carefully parsed, the addition is not a ringing endorsement as it appears on the surface.
What's worse? Hating people you have never met or idolizing people you have never met? Maybe spend your energy elsewhere instead of getting so worked up that you have to start calling people names.
Consider that your perspective on the man is hopelessly warped by the media frenzy surrounding him. Sadly, you are a victim of it just as much as he is.
Back in the day Reed told me that he added the leak detection to Purify v1 towards the end of the dev cycle and was very surprised that it wound up being one of their biggest selling points
I remember using that and libefence in the 90s to find memory leaks in my MUD. Purify was better. Easier to use and found more stuff. It is 2022, leaking memory still a problem.
Recently I was thinking about how the Netflix software strategy of hiring only experienced engineers [0] (and paying them really well), is pretty beneficial for almost any tech team/company.
Things like the CS experience that Reed Hastings has, seems like the key driver here.
I hadn't realized Pure and Netflix had the same founder; Purify was an amazing tool, which we had a license for HP-UX at uni. It was the kind of software where you knew you'd ask for it to be bought by any future employer.
A major difference is that purify is a binary transformation performed before or after linking. Valgrind/memcheck is a virtual machine that jit-compiles executable code and adds its instrumentation in the process. So, with purify, your linked executable includes the checks. With valgrind, the unmodified executable can run with or without checks.
The takeout is: valgrind is free and doesn't require recompilation, purify has a more complete UI, they both have similar performance impact and bug-finding abilities.
I used Purify back in the day, and while it worked well, getting it working was a royal PITA. IIRC it was necessary to recompile EVERYTHING with Purify.
That's not how it usually happens in my experience. In this specific case, I worked at Pure Atria back in the day and it was my impression that Reed had some hacker chops.
> EDIT: Wow. OK, I guess I'm the only one in the world who's ever seen that happen.
It happens regularly in larger companies. Most patents that I have also have others on them, because secondary people had to approve my applications. That's just how these things work. The only exceptions to this are those patent applications that I approved for others.
Remember, work on a patent may be more complex than the work on a corresponding invention. Also, the patent must be useful to the company in order for them to cover the bill. Taking a single employee's word for that is not enough.
The other people are on there because by law they have to be. Anyone involved in the invention in any way has to be on the patent or it can be found invalid.
It’s worth questioning but he wasn’t a big deal in the early 90s and if Wikipedia is correct he finished a Stanford CS MS in B 1988. This seems very much in character.
This is different from scientific papers, where it is poor taste to add people that did not actually contribute anything, but has no legal implications.
He co-founded Netflix in 1997.
https://en.wikipedia.org/wiki/Reed_Hastings