4 ms·

It's definitely odd. I'm inclined to attribute it to carelessness, and just not really thinking about the implications of precisely how they chose to implement

It's definitely odd. I'm inclined to attribute it to carelessness, and just not really thinking about the implications of precisely how they chose to implement the Rack specification. What's frustrating is that apparently there wasn't any performance testing of the 2.3 release that ought to have found this problem. It's been in the code base for about a year at this point.

dasil003 17y ago
How often is ActionResponse.body a String? It's not really clear to me without digging deeper, but I could see this easily being an edge case related to certain caching and rendering methods. In any case, suggesting that any performance testing of Rails would have magically found this problem is unfair. Many many many Rails app instances are serving orders of magnitude more than 10reqs/sec, so clearly this is not a universal problem. My guess as to the reasoning here is so that large files can be served without blocking while the whole thing is written to the socket. Granted, that's just my gut instinct looking at that code, maybe it is just careless...
- EvilTrout 17y ago
  I've been running Rails 2.3 in production for a while with great performance. In the article the author mentions it has to do with serving static files, which is something that nobody should be using Rails to do anyway. I'm not surprised nobody found it until now.
  - kscaldef 17y ago
    No, we only incidentally found it while working on a piece of our app that serves static files. The issue looks universal to me, although with dynamic content or smaller output, the relative hit would be less noticable.
    - EvilTrout 17y ago
      I stand corrected!
- kscaldef 17y ago
  I checked that, and I couldn't find any case where the body _wasn't_ a String. It's true, that the problem only becomes really glaring when the pages being served are quite large (this example was 1300 lines of HTML), but I still assert that good performance regression testing ought to have popped this up. With 100 lines of output, which is pretty reasonable, you'd still be looking at an extra 10ms on each request.

dasil003 17y ago

How often is ActionResponse.body a String? It's not really clear to me without digging deeper, but I could see this easily being an edge case related to certain caching and rendering methods. In any case, suggesting that any performance testing of Rails would have magically found this problem is unfair. Many many many Rails app instances are serving orders of magnitude more than 10reqs/sec, so clearly this is not a universal problem. My guess as to the reasoning here is so that large files can be served without blocking while the whole thing is written to the socket. Granted, that's just my gut instinct looking at that code, maybe it is just careless...

EvilTrout 17y ago
I've been running Rails 2.3 in production for a while with great performance. In the article the author mentions it has to do with serving static files, which is something that nobody should be using Rails to do anyway. I'm not surprised nobody found it until now.
- kscaldef 17y ago
  No, we only incidentally found it while working on a piece of our app that serves static files. The issue looks universal to me, although with dynamic content or smaller output, the relative hit would be less noticable.
  - EvilTrout 17y ago
    I stand corrected!
kscaldef 17y ago
I checked that, and I couldn't find any case where the body _wasn't_ a String. It's true, that the problem only becomes really glaring when the pages being served are quite large (this example was 1300 lines of HTML), but I still assert that good performance regression testing ought to have popped this up. With 100 lines of output, which is pretty reasonable, you'd still be looking at an extra 10ms on each request.