Re: Request consideration of DBD::Neo4p registration

On 13/02/14 09:24, Tim Bunce wrote:
> On Mon, Feb 10, 2014 at 08:59:17AM +0000, Martin J. Evans wrote:
>>
>>> That lets you, and others, implement other transport/connection modules.
>>> I mention this because LWP is not the fastest HTTP interface. There are
>>> several transports for Elasticsearch that are significantly faster.
>>> For example https://metacpan.org/pod/Elasticsearch::Cxn::NetCurl
>>
>> We found curl much faster than LWP - see
>> http://www.martin-evans.me.uk/node/117 for some numbers and a problem
>> I hit (and got around) with POSTing in Curl.
>
> That's a great article. Thanks.
>
> Any chance you could update those benchmarks Martin?  Would be nice to
> get numbers for the current versions and also include Hijk.
>
> I'd also be really interested in the results for HTTP::Tiny with
> Clinton's persistent connections pull request applied:
> https://github.com/chansen/p5-http-tiny/pull/30
> (I think that might get merged during the QA Hackathon in March).
>
> Tim.
> .
>

See http://www.martin-evans.me.uk/node/117#comment-2129

I did not find keep_alive = 1 made any difference to me but see my notes in the comment above.

Martin