switch to room list switch to menu My folders
Go to page: First ... 12 13 14 15 [16] 17 18 19 20 ... Last
[#] Thu Mar 04 2010 16:13:05 EST from Peter Pulse @ Uncensored

[Reply] [ReplyQuoted] [Headers] [Print]

Yea, Netware rocked.

[#] Fri Mar 05 2010 10:16:33 EST from fleeb @ Uncensored

[Reply] [ReplyQuoted] [Headers] [Print]

I was *just* thinking about Netware lately. I kinda thought they already stopped supporting it. Remarkable that they did continue to support it all this time.

But, it's not surprising that they've killed it now. Kind of a pity... it's a pretty decent operating system, for what it does.

[#] Fri Mar 05 2010 10:16:59 EST from Spell Binder @ Uncensored

[Reply] [ReplyQuoted] [Headers] [Print]

I've still got a copy of NetWare sitting around since I used to test IPX support for my company's switches.

I've always wondered why IPX never got more of a hold in the networking industry.
I always thought that IPX's explicit separation of the network and host addresses was a Good Thing(tm), though forcing the host address to be the host's MAC address was a pain in the butt. Although now IPv6 is doing pretty much the same thing.

Hmmm...SPX over IPv6?
Netware Binder

[#] Fri Mar 05 2010 11:38:18 EST from IGnatius T Foobar @ Uncensored

[Reply] [ReplyQuoted] [Headers] [Print]

IPv6 reminds me a lot of IPX, actually. They really do recommend that you let your hosts auto-configure, in a mode which uses a /64 subnet mask for every network, and the host portion of the address is derived from the MAC address.

All we need is a good stable way of letting the host register itself with a nameserver.

[#] Fri Mar 05 2010 11:59:50 EST from Spell Binder @ Uncensored

[Reply] [ReplyQuoted] [Headers] [Print]

I take it DDNS doesn't work as well as advertised?

[#] Fri Mar 05 2010 12:09:34 EST from IGnatius T Foobar @ Uncensored

[Reply] [ReplyQuoted] [Headers] [Print]

It works, but you have to

* Get people used to relying on it, even for servers

* Arrange for things like firewalls and VPNs to work with it

* etc

[#] Sat Mar 06 2010 08:04:34 EST from dothebart @ Uncensored

[Reply] [ReplyQuoted] [Headers] [Print]

ipx isn't iso/osi layered, for which reason you can't do switching and gatewaying it in a good way..

[#] Sun Mar 07 2010 07:29:34 EST from IGnatius T Foobar @ Uncensored

[Reply] [ReplyQuoted] [Headers] [Print]

How so? It follows the rules for layering about as well as IP does. 802.3 or whatever is Layer 2 when running on Ethernet; IPX is Layer 3; SPX is Layer 4; RIP or NLSP provide a routing table, while SAP or NDS provide service location.

To be honest, a properly configured IPX network "just works" without a lot of manual configuration, once you've got your directory and network numbers set up. Adding new hosts is just a matter of plugging them in and naming them.

It is my hope that IPv6 will bring some of that back. Plug in a host, let it autoconfigure an IP address, and then let it register with a name service.
With any luck, deliberately knowing and configuring your host's IP address will become a thing of the past.

[#] Mon Mar 08 2010 16:25:39 EST from Spell Binder @ Uncensored

[Reply] [ReplyQuoted] [Headers] [Print]

I'm with ya, IG, but the one concern I have is crufty code that isn't designed to handle an interface address change while there are active connections.

Or does DHCP now have an option for unlimited lease times?

[#] Mon Mar 08 2010 18:38:14 EST from IGnatius T Foobar @ Uncensored

[Reply] [ReplyQuoted] [Headers] [Print]

In IPv6 it's less common to use DHCP and more common to do stateless autoconfiguration -- the IPv6 address is determined by the host's MAC address. It isn't going to change unless you change the NIC, in which case you're not going to have any active connections.

[#] Mon Mar 08 2010 22:52:40 EST from Animal @ Uncensored

[Reply] [ReplyQuoted] [Headers] [Print]

why would you /want/ to have an unlimited lease?

[#] Mon Mar 08 2010 22:57:22 EST from fleeb @ Uncensored

[Reply] [ReplyQuoted] [Headers] [Print]

Unlimited lease means less network traffic.  heh.

[#] Tue Mar 09 2010 01:59:55 EST from Animal @ Uncensored

[Reply] [ReplyQuoted] [Headers] [Print]

something tells me DHCP traffic is dwarfed by youtube, streaming music,
and internet porn.

[#] Tue Mar 09 2010 05:49:37 EST from fleeb @ Uncensored

[Reply] [ReplyQuoted] [Headers] [Print]

Maybe a little.

[#] Tue Mar 09 2010 10:48:41 EST from Spell Binder @ Uncensored

[Reply] [ReplyQuoted] [Headers] [Print]

Re: Unlimited lease.

It would be ideal for servers. As IG mentioned, instead of having to statically configure IP addresses for all your server machines, it would be much easier to have the boxes auto-configure their IP addresses. For a server, though, you usually want that address to remain the same for as long as possible.
DHCP's notion of leasing IP addresses runs counter to that principle.

As IG pointed out, though, IPv6 allows hosts to auto-configure their interface addresses without the need of a DHCP server. Plus, IPv6's auto-configuration is predictable: i.e., a host will get the same address every time.
IPv6 Binder

[#] Tue Mar 09 2010 11:16:13 EST from Peter Pulse @ Uncensored

[Reply] [ReplyQuoted] [Headers] [Print]

DHCP or something like it makes sense because it allows efficient use of the address space. ipv6 is not very clever, it just throws bits at the problem, greatly increasing the size of the header on every ip packet, everywhere.
Then, having added all that overhead, encouraging people to waste most of those bits by assigning a huge local address range.... Seems stupid to me.
But maybe I am missing some clever aspect of it.

[#] Tue Mar 09 2010 23:44:09 EST from IGnatius T Foobar @ Uncensored

[Reply] [ReplyQuoted] [Headers] [Print]

IPv6 solves the problem of lazy network administrators who size their subnets based on administrative convenience rather than actual requirements -- i.e. "all the world's a /24." As anyone who's worked at an ISP can tell you, it's quite annoying to fend off subscribers who deman "a Class C" for no other reason than because they want everything to fall on an octet boundary. Some will go so far as to falsify their documentation and claim that they're rapidly heading for 250 hosts when they've got half a dozen servers in their rack.

In IPv6, the actual recommended practice is "all the world's a /64." Including all those bits in every packet is nothing new -- IPX handled it just fine, with 80-bit source and destination addresses in every single packet, on much less powerful network hardware than we have today.

I think it's really based on pragmatism rather than finding the technologically optimal solution. The IETF knows that there are an awful lot of careless and stupid network administrators out there.

[#] Wed Mar 10 2010 14:32:21 EST from Spell Binder @ Uncensored

[Reply] [ReplyQuoted] [Headers] [Print]

And, frankly, many modern routers can already forward IPv6 datagrams at line rate. One downside is that, due to the increased header size, an IPv4 datagram that may have fit into a 64-byte Ethernet frame now may not due to the extra 20-bytes of IP header. As link speeds continue to increase, though, that 20-byte difference won't matter. On a 1 Mbps link, 20 bytes takes 160 microseconds (usec) to transmit. Even for delay-sensitive applications like video conference or ciruit-emulation services, that's well within the acceptable margins for end-to-end delay.

[#] Thu Mar 11 2010 08:00:51 EST from IGnatius T Foobar @ Uncensored

[Reply] [ReplyQuoted] [Headers] [Print]

This is kind of a neat story.

30 years of the IEEE 802 committee. This is a nice little history of some of the early days of computer networking, and what happened along the way.
(full disclosure: I only skimmed it; one of the reasons I'm posting it here is to remind myself to read it in full when I have time)

[#] Thu Mar 11 2010 12:15:20 EST from Ford II @ Uncensored

[Reply] [ReplyQuoted] [Headers] [Print]

a name service. With any luck, deliberately knowing and configuring
your host's IP address will become a thing of the past.

That's probably why they made ipv6 numbers so big no human could readily use them the way we currently pandy about with ipv4 addresses.

Go to page: First ... 12 13 14 15 [16] 17 18 19 20 ... Last