When the -M option is used, bgpq3 is significantly slower than without it. This even happens when building non-JUNOS filters, where -M actually becomes a no-op (i.e., the output is the same with or without it), as demonstrated here:
tore@echo:~/git/bgpq3$ time bgpq3 AS-TELENOR | md5sum
578aff1c367b9d96fa0b57a4daafc709 -
real 0m0.847s
user 0m0.007s
sys 0m0.005s
tore@echo:~/git/bgpq3$ time bgpq3 -M foo AS-TELENOR | md5sum
578aff1c367b9d96fa0b57a4daafc709 -
real 0m23.427s
user 0m0.009s
sys 0m0.021s
It seems odd to me that the -M option should have such a big effect on the program, after all it's just a string that should be included in Juniper-specific output, so the way I see it it shouldn't cause any additional heavy processing, network, traffic or anything else that could conceivably explain the observed slowdown.
When the -M option is used, bgpq3 is significantly slower than without it. This even happens when building non-JUNOS filters, where -M actually becomes a no-op (i.e., the output is the same with or without it), as demonstrated here:
It seems odd to me that the -M option should have such a big effect on the program, after all it's just a string that should be included in Juniper-specific output, so the way I see it it shouldn't cause any additional heavy processing, network, traffic or anything else that could conceivably explain the observed slowdown.