RFR: 6968351: httpserver clashes with delayed TCP ACKs for low Content-Length

[Daniel Jeliński]

On Fri, 19 Apr 2024 10:48:41 GMT, Daniel Fuchs <dfuchs at openjdk.org> wrote:

>> src/jdk.httpserver/share/classes/sun/net/httpserver/ExchangeImpl.java line 210:
>> 
>>> 208:         this.rcode = rCode;
>>> 209:         String statusLine = "HTTP/1.1 "+rCode+Code.msg(rCode)+"\r\n";
>>> 210:         ByteArrayOutputStream tmpout = new ByteArrayOutputStream();
>> 
>> Wouldn't it be better to continue using a BufferedOutputStream here? That would put a limit on how much data would be buffered when writing the headers, in case of long response headers.
>
> Forget I asked this. @djelinski pointed out this is necessary to avoid having to flush at line 281. But I see that at line 686 you replaced rawout with a BufferedOutputStream - so do we still need the temporary ByteArrayOutputStream? Can't we just write directly to `ros` (possibly assert ros instanceof BufferedOutputStream if we want to make sure...)

Before doing that, you might want to check the HTTPS case. As far as I can tell, we don't do any buffering when SSL is in use.

-------------

PR Review Comment: https://git.openjdk.org/jdk/pull/18667#discussion_r1572303202