[Replicant] [PATCH v2] Fix "RepWifi connection not correctly reported to userspace apps" Issue #1792.

Tj replicant at iam.tj
Tue Aug 8 09:20:43 UTC 2017


On 08/08/17 00:47, Fil wrote:
> 
> Joonas Kylmälä wrote:
>> Think about the 1% who do not use NAT (actually in Finland the ISPs give
>> usually 5 IPv4 addresses so it might not be too uncommon). You mentioned
>> earlier the link local addresses so we could make a list of similar
>> non-working IPv4 addresses (in case there are others) and then check if
>> the address belongs to that list and act accordingly. The function
>> isLinkLocalAddress() can be used.
> 
> I didn't know about the situation in Finland. I'm still doubtful about
> those public addresses being actually assigned directly to your WLAN
> interface, but I guess I don't know enough to assume they're not. Let's
> say we need to consider that 1% chance too..

This wasn't too uncommon in other countries in the 1990s amongst
'professional/business' ISPs until it became clear the IPv4 address
space was becoming scarce.

What usually happened is the ISP allocated a /29 sub-net:

255.255.255.255 = /32 = 1 address
255.255.255.254 = /31 = 2 addresses
255.255.255.253 = /30 = 4 addresses
255.255.255.252 = /29 = 8 addresses

To further complicate matters usually the CPE router would be allocated
a single /32 IP address from another range entirely. It would be the
responsibility of the customer to configure and allocate the /29 sub-net
either statically or using DHCP. The default route would be the IP
address of the CPE router, so:

A <Internet>      0.0.0.0/0
B <ISP gateway>   a.b.c.d/32 default via dev A, i.j.k.l/29 via e.f.g.h
dev C
C <CPE gateway>   e.f.g.h/32 default via a.b.c.d, i.j.k.l/29 via dev D
D <LAN PCs>       i.j.k.l/29 default via e.f.g.h

Of the 8 addresses only 5 are usable for routing because:

0 = i.j.k.l = sub-net address
1 = i.j.k.l+1 = usually the CPE router's LAN address
2 = i.j.k.l+2 }
...           }- allocated to devices on the LAN
6 = i.j.k.l+6 }
7 = i.j.k.l+7 = sub-net broadcast address


More information about the Replicant mailing list