Summary:	ASTERISK-08627: Installer Static IP check 'fails' valid IP.
Reporter:	Simon Wood (mungewell)	Labels:
Date Opened:	2007-01-20 23:48:36.000-0600	Date Closed:	2007-01-22 11:05:18.000-0600
Priority:	Minor	Regression?	No
Status:	Closed/Complete	Components:	Core/General
Versions:		Frequency of Occurrence
Related Issues:
Environment:		Attachments:
Description:	Hi, whilst installing *NOW 1.4beta2 I attempted to set a static IP of 10.38.39.0/netmask 255.0.0.0. This was reported as an invalid IP, which it isn't. Changing IP to 10.38.39.1 allowed install to continue. Simon.
Comments:	By: mhardeman (mhardeman) 2007-01-22 00:09:21.000-0600 This issue is so bizarre I wanted to comment :-) Out of morbid curiosity, does anyone actually ever assign .0's as an endpoint IP address? While I recognize that the particular IP and netmask combination mentioned in the description is technically valid, it would certainly preclude the later use of class C style subnetting of the 10.* space and would prohibit addressing IPs in the 10.38.39.* network in an "extruded subnet" sense. Further, it occurs to me that no one giving any real thought to the matter really ever wants to have an IP broadcast domain as large as an entire class A space. If you have need of a private IP space as large as that, chances are that it spans many sites and there are many routers involved.  A broadcast domain that large just makes no sense and I suspect would scale rather miserably, next to other options. Of course, in time, you would discover this. The neat part is, with the advent of modern managed switches and nice routers, there would be an easy fix -- properly subnetting at some point in the future... Until...  You guessed it...  (Or perhaps not) You discover that you have clients that prohibit you from breaking up the space into nice boundaries.  (This is not entirely true, as you could go to larger blocks than a /24 and still have that .0 abomination be valid -- but lets be real here -- that's sort of ridiculous.) While I agree that a strict interpretation of the rules of classless IP addressing may allow for IP stations with .0 addresses in some cases, the practice seems quite uncommon and almost always represents a serious network design mistake.  (Some exceptions, but really rare cases.) That said, perhaps the "kindest" behavior of any IP application is to act with extreme prejudice toward such blasphemous indifference to IP addressing as preached by the original IP addressing gods. My vote (for all that it's worth: i.e. not much) is that any IP address ending in .0 (or .255 for that matter) is unfit for use by an IP endpoint and that the user should in fact receive an especially chastising error message to discourage further perverse behavior. LOL, please forgive this (I hope you all realize) terribly tongue-in-cheek post.  I just have a special hatred for ISPs and other network infrastructure types who would have us believe that .0 and .255 are OK for use now (in large block cases) when in fact so many legacy applications fail to properly handle the case.  And also certain cable providers who wheel-and-deal to get previously bogon listed IP space allocated to them.  *sigh* slews of firewall updates... Seriously though - in a world of so many legacy devices and so much legacy code held over (especially in embedded devices -- think IP phones with questionable IP stacks: looking at you, c15c0) -- one would probably think that it is especially improper to install a machine delegated to a SERVER role with a .0. By: James Lyons (james) 2007-01-22 11:05:18.000-0600 ^