| |||||||||||||
RustRustPosted Apr 4, 2013 20:29 UTC (Thu) by khim (subscriber, #9252)In reply to: Rust by ncm Parent article: Mozilla and Samsung building a new browser engine
If (as it appears) Rust is as expressive as C++, equally fast, and link-compatible with C, it has a unique opportunity because it's much, much simpler than C++. Nope. There are fundamental difference between these. To go from C to Java you need to jump over 100 feet high wall and to go from C to Haskell you need to jump over 1000 feet high wall. With Rust this wall may be 50 feet high or may be even 10 feet high, but it's still a wall and it'll keep users from defecting from C camp. C++? For C++ it's kinda 500 feet-high staircase: sure, the end position it pretty far removed from where you've started but you can go there at the pace you want. You can use as many or as few C++ features as you want - which means that any C programmer can become C++ programmer by learning about couple of things s/he wants/needs. Sure, s/he may not be very efficient C++ programmer, but s/he'll be a C++ programmer even the only thing s/he uses are mutex classes or something like this. Eventually C++ can replace C more-or-less everywhere (the last holdouts like binutils and gcc are slowly switching to C++ camp). Well, may be not kernel (because it uses C++-incompatible dialect of C), but even that may happen in 10 years or so. If you introduce incompatible syntax, incompatible paradigms (and insist that said paradigms must be used from the start) then substantial percentage of developers will just be kept behind the wall such changes reproduce. Rust may succeed as a language or it may fail but it can not be next C++. It's not in the cards. No language is forever. Rust might not be the true Glorious Successor to C++. If not, whatever is will look a lot like Rust. I seriously doubt it. Go, Rust and other "modern" languages is not designed to be C replacement and this means that at best they'll coexist with C/C++ (at worst they'll go to the dustbin of history). Think another, parallel world of Pascal: it was quite popular till a few years ago but it's nominal successor never actually caught up (as successor!) because it was incompatible. Rust is in similar position and it's not even named as "C successor" by people who designed C... Or another data point: Fortran. The last version which was actually incompatible with it's predecessor to significant degree was FORTRAN 77 - and it was almost enough to ruin the transition. Planned transition governed by most compiler vendors! And now you think that another NCP style transitions will happen? Dream on.
(Log in to post comments)
Rust Posted Apr 5, 2013 7:00 UTC (Fri) by ncm (subscriber, #165) [Link]
Fortran is an excellent example. Subsequent Fortran releases are more or less backward-compatible, and recent ones even support pointers and recursion, but how many places are they used? When starting a project where Fortran would have been required, people use C++ now because it's faster and more expressive. C++ has turned out to be the Glorious Successor to Fortran, despite being just barely link compatible, in the same way that FORTRAN succeeded assembly language.
The GStC++ will not be anything that is obliged to carry forward C++'s avoidable complexities (largely, forced C compatibility, name lookup horrors). People using C++ now will continue using it, but in maybe ten years new organizations and projects that need C++'s power will begin to pick up Rust (or something) instead. Second-tier languages will come and go, and be used for the easy problems, as always.
My point is that Rust is the first language above the horizon that seems physically capable of supplanting C++ on problems where, before, nothing else could even be seriously considered. If it sputters out, the world will continue with C++ as before until the next contender matures.
Rust Posted Apr 5, 2013 7:32 UTC (Fri) by khim (subscriber, #9252) [Link] Subsequent Fortran releases are more or less backward-compatible, and recent ones even support pointers and recursion, but how many places are they used? In more places then you expect. I know few people who thought like you do till they actually participated in bioresearch, weather prediction research and so on. They found out to their sudden surprise that new projects are still written in Fortran. Usually it's a mix: IT guys write the supplemental logic in C/C++ and non-IT scientists write their algorithms in Fortran. In that sense Fortran is still alive. Number one request for Exacycle is Fortran support (perhaps they already implemented Fortran support at this stage? I don't know). People using C++ now will continue using it, but in maybe ten years new organizations and projects that need C++'s power will begin to pick up Rust (or something) instead. Second-tier languages will come and go, and be used for the easy problems, as always. And Rust in one of these second-tier languages. The same happens with other industries, too. Think track gauges or QWERTY keyboard: once upon time it was easy to change track gauges or layout of keyboard but by now it's basically impossible. You can change some auxiliary pieces (railroad ties are made today from stone, wood, plastic or even fiberglass) and specialized cases use specialized gauges or keyboards, but mainstream is stuck. My point is that Rust is the first language above the horizon that seems physically capable of supplanting C++ on problems where, before, nothing else could even be seriously considered. Nope. Rust is yet-another-pretender with overblown hype. If it sputters out, the world will continue with C++ as before until the next contender matures. Well, the next contender will probably be called C++17, so it's not like we'll need to wait too long.
Rust Posted Apr 5, 2013 8:38 UTC (Fri) by ncm (subscriber, #165) [Link]
Yes, Fortran is still used heavily in certain domains, but no longer because it is, as it once was, the only reasonable choice. Nowadays its continued use is more a matter of inertia than of technical merit.
You seem to be asserting that C++ will always be the only reasonable choice for those difficult problems where there is, at this moment, no other. People used to say that about Fortran. They were wrong too.
That's not to say that Rust (or its descendant, TiN?) really will be the new heavy. What is certain is that there will someday be a new heavy that is not C++XX. Furthermore, it will differ from C++ in many of the ways that Rust does.
Rust Posted Apr 5, 2013 15:00 UTC (Fri) by tjc (subscriber, #137) [Link]
I work in bioresearch, and we use Java for almost everything. That's not the decision I would have made, but that's the trend. Nearly all the organizations we work with use Java, so it's hard to resist.
We used to use C and Perl, which was a common combination in bioinformatics, but that's been fading away -- especially C, I am sorry to say.
|
|||||||||||||