[
https://issues.apache.org/jira/browse/SOLR-7442?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14506295#comment-14506295
]
Shawn Heisey commented on SOLR-7442:
------------------------------------
HTTP/2 support would be awesome. Incredibly awesome. I'm absolutely sure it
would be more efficient. I read the entire draft RFC, and I'm *really* excited
about the technology and how it is going to improve performance on the world
wide web.
I do not think, however, that most users would see an *ENORMOUS* efficiency
increase from HTTP/2, particularly if they are on a switched LAN that's gigabit
or faster. If you're accessing your Solr instance across a high-latency
network like the Internet or a private WAN, then you probably would see a
noticeable performance boost ... but I don't think this describes most users.
If switching http libraries is a significant amount of change to the codebase,
which I suspect it would be, I don't think HTTP/2 is enough of a reason to
endure the pain. We should only switch if there are significant advantages in
other areas, such as ease of use, stability, a significant jump in LAN-based
performance benchmarks, etc.
I could be wrong about how invasive a change in libraries is, of course. If
it's not a major pain, let's give it a try and see whether there is any
verifiable advantage on a well-configured LAN.
> HttpClient replacement for HTTP/2.0
> -----------------------------------
>
> Key: SOLR-7442
> URL: https://issues.apache.org/jira/browse/SOLR-7442
> Project: Solr
> Issue Type: New Feature
> Reporter: Ishan Chattopadhyaya
>
> Given that it would take about a year for Apache HC's HttpClient to support
> HTTP/2 (as per [~olegk] in SOLR-6865), adding this issue for exploring the
> way forward. It would be preferable to move to HTTP/2 and be able to use
> features like async http calls. Maybe we could look for alternatives to HC
> HttpClient, (e.g. Jetty's HttpClient)?
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
