Difference between curl expressions - linux

Question

I have an API server running at localhost:3000 and I am trying to query it using these two expressions:
[wani#lenovo ilparser-docker]$ time (curl "localhost:3000/parse?lang=hin&data=देश" )
{"tokenizer":"<Sentence id=\"1\">\n1\tदेश\tunk\n<\/Sentence>\n"}
real 0m0.023s
user 0m0.009s
sys 0m0.004s
[wani#lenovo ilparser-docker]$ time (curl -XGET localhost:3000/parse -F lang=hin -F data="देश" )
{"tokenizer":"<Sentence id=\"1\">\n1\tदेश\tunk\n<\/Sentence>\n"}
real 0m1.101s
user 0m0.020s
sys 0m0.070s
Why does the second expression take so much more time?
With more verbosity:
[wani#lenovo ilparser-docker]$ time curl -v localhost:3000/parse -F lang=hin -F data="देश"
* Trying 127.0.0.1...
* Connected to localhost (127.0.0.1) port 3000 (#0)
* Initializing NSS with certpath: sql:/etc/pki/nssdb
> POST /parse HTTP/1.1
> Host: localhost:3000
> User-Agent: curl/7.43.0
> Accept: */*
> Content-Length: 244
> Expect: 100-continue
> Content-Type: multipart/form-data; boundary=------------------------1eb5e5991b976cb1
>
* Done waiting for 100-continue
< HTTP/1.1 200 OK
< Content-Length: 70
< Server: Mojolicious (Perl)
< Content-Type: application/json;charset=UTF-8
< Date: Mon, 21 Mar 2016 11:06:09 GMT
< Connection: keep-alive
<
* Connection #0 to host localhost left intact
{"tokenizer":"<Sentence id=\"1\">\n1\tदेश\tunk\n<\/Sentence>\n"}
real 0m1.106s
user 0m0.027s
sys 0m0.068s
[wani#lenovo ilparser-docker]$ time curl -v localhost:3000/parse --data lang=hin --data data="देश"
* Trying 127.0.0.1...
* Connected to localhost (127.0.0.1) port 3000 (#0)
> POST /parse HTTP/1.1
> Host: localhost:3000
> User-Agent: curl/7.43.0
> Accept: */*
> Content-Length: 23
> Content-Type: application/x-www-form-urlencoded
>
* upload completely sent off: 23 out of 23 bytes
< HTTP/1.1 200 OK
< Server: Mojolicious (Perl)
< Content-Length: 70
< Connection: keep-alive
< Date: Mon, 21 Mar 2016 11:06:24 GMT
< Content-Type: application/json;charset=UTF-8
<
* Connection #0 to host localhost left intact
{"tokenizer":"<Sentence id=\"1\">\n1\tदेश\tunk\n<\/Sentence>\n"}
real 0m0.031s
user 0m0.011s
sys 0m0.003s
Expect: 100-continue sounded fishy, so I cleared that header:
[wani#lenovo ilparser-docker]$ time curl -v -F lang=hin -F data="देश" "localhost:3000/parse" -H Expect: --trace-time
16:48:04.513691 * Trying 127.0.0.1...
16:48:04.513933 * Connected to localhost (127.0.0.1) port 3000 (#0)
16:48:04.514083 * Initializing NSS with certpath: sql:/etc/pki/nssdb
16:48:04.610095 > POST /parse HTTP/1.1
16:48:04.610095 > Host: localhost:3000
16:48:04.610095 > User-Agent: curl/7.43.0
16:48:04.610095 > Accept: */*
16:48:04.610095 > Content-Length: 244
16:48:04.610095 > Content-Type: multipart/form-data; boundary=------------------------24f30647b16ba82d
16:48:04.610095 >
16:48:04.618107 < HTTP/1.1 200 OK
16:48:04.618194 < Content-Length: 70
16:48:04.618249 < Server: Mojolicious (Perl)
16:48:04.618306 < Content-Type: application/json;charset=UTF-8
16:48:04.618370 < Date: Mon, 21 Mar 2016 11:18:04 GMT
16:48:04.618430 < Connection: keep-alive
16:48:04.618492 <
16:48:04.618590 * Connection #0 to host localhost left intact
{"tokenizer":"<Sentence id=\"1\">\n1\tदेश\tunk\n<\/Sentence>\n"}
real 0m0.117s
user 0m0.023s
sys 0m0.082s
Now the only time taking thing left is: Initializing NSS with certpath: sql:/etc/pki/nssdb. Why does curl do that in this context?

Answer 1

After a little help on IRC from #DanielStenberg, I came to know that the db load is present because curl inits nss in that case since curl needs a good random source for the boundary separator used for -F . Curl could have used getrandom() syscall or read bits out of /dev/urandom since boundary separators don't need to be cryptographically secure in any way, but curl just wants secure random in some other places so curl reuses the random function that it already has.