Devconf 2016: Interview with the developer SphinxSearch

We love doing interviews with our speakers. This time - an interview with Andrei shodan Aksenov , the creator of the search engine SphinxSearch (I don’t know how it is now, but earlier Habr used this particular engine to search the site) and, as it turned out, a couple more engines. At our conference, he will talk on the eternal topic: Devs vs CTO vs CEO . The interview itself turned out quite interesting. Andrew honestly and frankly answered questions, for which he had great respect :)

I know you as a developer of Sphinxsearch, but as your brief biography, which you gave in the report description, shows, computer graphics seem to be your main passion?

Everything is both simpler and harder at the same time, there is no main mega-growth :) There are some interesting areas, there are indifferent ones. Do any low level, such as the search engine, 3D games, video capture processing, a tiny under-hypervisor (V8086), patch MySQL or the operating system kernel, write special modules to Apache, improve the well-known compression algorithm, or even just tinker with particularly thick and inhibiting SQL queries - all this it is quite interesting (and sometimes not! but more often interesting). But it is not important to draw molds for GUI, on the trendy Node.js or dead Delphi, or to edit gloomy reports in 1C - not very good. I speak from personal experience, tk. to some extent engaged in all of the above and more.

In general, I googled your biography and found nothing. Even Wikipedia articles about you not!

And this is very good, I am a secretive and unsociable person. Yes, and achievements on the article (pun desired) yet, I hope, do not pull, muhaha.

Fill the gap. Tell me why and how? About TheEngine . Tell me why TheEngine ended and how Sphinx started.

All simple. Boris Batkin, with whom we met at FIDO, started his game development company and suddenly called me back there, well, I just as suddenly agreed and went. Skyfallen has managed to build a 3D engine, a toolbox and several games on it, plus licensing the technology for several more games. At first I was responsible for the actual graphics, animations and other low levels, then gradually took over the overall leadership of the development team. And then, as it were, he got bored and soooely retired. (Fun fact: the first time I personally saw Boreas about 7m or something, a year of acquaintance and 4th year of my work. And it happens!)
')
Sphinx started even before this, the very first version is dated something like September 2001. There is a formal date of birth, that is, the day when for 30kb sources in the daddy with the name dbindex it was the name Sphinx that was thought up. But I have never celebrated it and forget it all the time. Then it lay down in the table of 5 year olds, then, under the influence of Peter Zaitsev, it was posted on the Internet, found its small user base and very tiny, but nonzero, ready to pay for any customers. Having combined this business for some time, I realized that I could no longer work distinctly in game devays during the day and on the search in the evenings and weekends, I have to choose something. Sphinx was painfully chosen, it was somewhere in 2007.

What's up with the sphinx now? The latest version is 2.2.10 (2015-09-09).

This is the last officially assembled package, not the latest version. There is version 2.3, fresher, lives in the master, it is quite supported by itself. There are bug fixes in 2.2, also rolled in on occasion. We support the 2.x branches to some extent and we are developing a large remake of 3.0 as much as we can, the first is immediately visible in the githabb commits, the second is not visible at all, but it will also be downloaded to the github sooner or later. However, we have had little strength lately for various reasons, so today, alas, it turns out like this, slowly and sadly. In particular, here on the packages podzabili. As usual in the open source, you can do better - come to visit us, do it well, we will be only happy :)

And finally, Unigine . What are you doing there?

I help the existing development team with the project and grow above myself. This includes everything from writing code to reading lectures.

In general, I have a standard stereotype, the graphics engine means for games, but it seems that Unigine is more used by projects from other industries (although games too). Is there any difference in the approaches, or the fact that “The Unigine engine is recognized by experts as the best 3D technology for Linux” (taken from the wiki) played its role?

In the niche of the so-called. professional simulations (these are simulators, architecture, VR and all that) have their own special requirements, which on one hand seem to be small, but as soon as you get busy, change everything. A classic example: in most games, a typical location has a size of, say, 1x1 km, or there are as many as 10x10 km, plus the accumulated errors are not very important. Therefore, to store any coordinates in the mass of places you can use a 32-bit float and everything is fine. And for simulators this is not enough: for example, at a distance of 3000 km the train will go through the whole of Australia, but it will not come there. Or, uh, the product will take and fly past, which is non-statutory behavior and therefore decidedly unacceptable. Oops, suddenly you need to store the coordinates in 64-bit double, and the choice of engines is sharply reduced. Or, here's another example: you can simply take and load a detailed model for 5 million polygons into the engine and it will heal, or you may not be able to do it at all, maybe. the engine will fall. Again, a rather interesting difference !!! But in games it is not necessary for anyone, so they don’t do much of it. So yes, Unigine is primarily about this, but for the purpose of building Unreal games or Unity, most likely, they will be better suited.

What is it like to be a developer of an open source engine such as Sphinx? What do you have to do?

What - difficult to answer. Well, uh, well, there is such a project in what would be achievements, well, uh, what else to say. Well, yes, well, Sphinx, well ... To be engaged in it at different points in time had to absolutely everyone in a row. When you write the code shock, when you talk to potential customers in the marathon mode (and you think: better code to pee), when you lie and spit on the ceiling, everything happened.

"You write engines for food" you say ... is it really that bad? Little buy support? Maybe you have a couple of mansions in Europe, and then you cry the blues? :)

And support for Sphinx has never been purchased in industrial quantities, and we have not learned how to sell it cleverly. Historically, at different moments, either naked Lucene, or Solr, or, today, Elastic, taxied, we were always such an alternative. Well, just like FreeBSD, someone still uses it. Although I personally, of course, prefer the Postgres analogy, about which in the distant years it was not heard at all, and then you didn’t die and even made you more popular lately.

There are no mansions anywhere, alas.

Want a new engine? In what field? What are you interested in now?

As always, something rather nontrivial is technically, with applied benefits, well, and that is understandable personally for my limited brain. Here it is right this second especially curious what is happening in the field of machine learning, for example. It used to be entertaining there, and in recent years quite interesting achievements will be able to classify objects in the pictures almost better than a person, then they will beat the world champion in Go - hearty !!! But I am not well acquainted with the region, and I consider this a flaw.

To report. Devs vs CTO vs CEO. Well, the developer, you, of course, was. And CTO, CEO? (Wouldn’t there be a one-sided report? CTO and CEO of SphinxSearch ... probably won’t work, it’s unlikely you had a permanent team there. Or was it?) And in general, tell me about your experience.

Of the Skyfallen companies mentioned - from programmer to CTO, Sphinx - all at once simultaneously and in one bottle, Unigine is some kind of mix of programmer duties and CTO, so I think the point of view has grown quite versatile.

What kind of report ? Training for managers? Anger management course for developers? Conciliatory mantras?

What kind of report - I do not know yet. The theme, as it happened, was partially forced, but close :) Therefore, I hope that something interesting will grow old.

Thank you shodan for the interview. By the way, come to Devconf 2016 . The program has already been formed. Soon - June 17th. There you can talk with him and make sure that his “I am a secretive and inhospitable person” is nonsense :)