Re: Replace history's "<remote>" with client's IP

[Derek Price]

Mark D. Baushke wrote:

>Adding that to the possibity of two decimal represenatations of a 16-bit
>port number means that there needs to be enough space in buf for 
>
>"xxxx:xxxx:xxxx:xxxx:xxxx:xxxx:ddd.ddd.ddd.ddd ddddd ddddd\0"
>
>which is 58 characters... allow a few extra for a nice round 64 number...
>  
>

Why not just Xasprintf the value? I don't see a good reason for using a
fixed length buffer here.

Also, there is no reason to reimplement the code to make the addr into a
string. The POSIX getnameinfo() function already exists to turn a struct
sockaddr * into a printable IP address or FQDN, as you prefer, and it
should already handle both IPV4 & IPV6. I think that is about as current
as it gets and its sister, getaddrinfo() is already in GNULIB, so we
could probably get a reimplementation of getnameinfo() into GNULIB too.
Of course, then you'd have to stick with the fixed length buffer. :)

Reverse lookup could pretty easily be made into a config option like it
is for Apache, since it is slow.

>From the getnameinfo() man page:

In order to assist the programmer in choosing reasonable sizes for the
supplied buffers, <netdb.h> defines the constants
# define NI_MAXHOST 1025
# define NI_MAXSERV 32
The former is the constant MAXDNAME in recent versions of BIND’s
<arpa/nameser.h> header file. The latter is a guess based on the ser-
vices listed in the current Assigned Numbers RFC.

Of course, those numbers are only relevant if you need the port number
or want a reverse lookup on the IP when possible, and the constants are
not portable according to POSIX, but they'd be easy to provide
substitutes for when not defined.

Regards,

Derek

-- 
Derek R. Price
CVS Solutions Architect
Ximbiot <http://ximbiot.com>
v: +1 717.579.6168
f: +1 717.234.3125
<mailto:derek@ximbiot.com>