JPKCom DB

IP Tools — Tips & Tricks

Tricks for the IP Tools: understanding CIDR in your head, subnetting strategy, ULA vs. GUA, common mistakes, and combining with other JPKCom tools.

Back to the overview: IP Tools · Open the tool live: www.jpkc.com/tools/ip/

The manual explains every tab, the examples show the calculations. This page is about the understanding behind them: how to read CIDR in your head, how to plan a network sensibly, when to use ULA over GUA, and which mistakes keep happening.

Understand CIDR instead of memorizing it

The prefix after the slash is just the number of network bits. Everything else follows from it:

  • Host bits = 32 − prefix, and therefore total addresses = 2 to the power of host bits. A /24 has 8 host bits → 256 addresses, a /26 has 6 → 64, a /30 has 2 → 4.
  • Usable hosts = total addresses − 2 (the network and broadcast addresses come off). Exceptions: a /31 has exactly 2 usable hosts per RFC 3021 (for point-to-point links), a /32 is a single host.
  • Each bit halves. From /24 to /25 the block halves (256 → 128), from /25 to /26 again (128 → 64). The other way, each bit fewer doubles it.

If you're unsure: the Binary Breakdown in the Subnet Calc tab colors network and host bits differently, so you literally see where the prefix ends. And the IPv4 CIDR Quick Reference in the Reference tab lists the mask and host count for every prefix; the Common button hides the exotic ones.

Subnetting strategy

  • Plan from the largest requirement. Count the hosts you need per segment, round up to the next power of two (plus the two reserved addresses), and pick the matching prefix. 100 hosts need a /25 (126 usable), not a /26 (62).
  • Use VLSM instead of cutting everything the same size. With Variable Length Subnet Masking you give large segments a /24 and point-to-point links a /30 (or /31). The Subnet Splits in the Subnet Calc tab show you directly which narrower prefixes a block divides into and how many sub-networks result.
  • Aggregate where you can. Several adjacent networks can often be summarized into a shorter prefix (supernetting). That's exactly what Range → CIDRs is for: enter the overall range and see how few blocks it takes to cover it — for IPv4 and IPv6 alike.
  • /31 for router links. On point-to-point links, a /31 (RFC 3021) needs half the addresses of a /30. The tool calculates both correctly.
  • Plan IPv6 on nibble boundaries. With IPv6 you don't count hosts, you count networks: every LAN segment gets a /64. Split a /48 or /56 at the /52, /56, /60, and /64 boundaries — each one moves by a whole hex digit, so the prefixes stay readable. Those are exactly the steps the Subnet Calc shows for an IPv6 prefix under Subnet Splits, along with the /64 subnets row.

ULA vs. GUA — which IPv6 address for what

  • GUA (Global Unicast, 2000::/3) are publicly routable — the normal case in the IPv6 internet. They usually arrive via prefix delegation from your provider (typically a /48 or /56).
  • ULA (Unique Local, fd00::/8) are the private counterpart, comparable to RFC 1918 addresses in IPv4. They aren't routed to the internet and suit internal services, lab and test networks, and addresses meant to stay stable independent of the ISP assignment.
  • The real ULA advantage is the random Global ID. Because the ULA generator draws the 40-bit Global ID cryptographically at random, your prefix is statistically unique worldwide. Two networks generated this way can be merged without an address conflict — unlike IPv4, where the same 192.168.0.0/16 collides everywhere.
  • Generate your ULA prefix once and keep it. The whole point is stability: roll a new one each time you need it and you lose exactly the uniqueness that makes ULA worthwhile. Generate the /48 once, document it, and split it into your /64 segments via the 16-bit Subnet ID.

Common mistakes

  • Assigning the network or broadcast address to a host. The first and last address of a subnet are reserved. The calculator shows you with First/Last Usable Host exactly the range you may actually hand out.
  • Entering an invalid "mask." A subnet mask has to be contiguous (255.255.255.192), not full of holes (255.0.255.0). For a non-contiguous mask the Subnet Calc tab reports "the 1-bits must be contiguous" — that's not a bug, it's correct behavior.
  • Leaving out the prefix. The subnet calculator and CIDR → Range always need a prefix (/24) or, for IPv4, a mask. The tool rejects a bare address instead of silently assuming a /24 — so you never end up calculating with the wrong network by accident.
  • Overlooking the host-bits note. Enter 10.0.5.7/16 and the tool calculates with the network 10.0.0.0/16, telling you so in the line "Host bits are set — the network is …". If that note shows up even though you meant to enter a network (in a firewall rule, say), you've probably made a typo.
  • Leading zeros in IPv4 addresses. 192.168.010.1 looks harmless, but some tools read 010 as octal 8. That's why the tool rejects such input and tells you the unambiguous spelling. When you see that message, also check the source you copied the address from.
  • Mixing up IPv6 and IPv4 when it comes to ranges. IPv6 has no broadcast — multicast (ff00::/8) takes over that job. That's why the subnet calculator shows First address and Last address for IPv6 instead of broadcast and usable hosts.
  • Using documentation ranges in production. 192.0.2.0/24, 198.51.100.0/24, 203.0.113.0/24, 2001:db8::/32, and the larger 3fff::/20 are explicitly reserved for examples. The Reference tab flags them — use them for documentation, never for real devices. The same goes for the benchmarking networks 198.18.0.0/15 and 2001:2::/48, which often get pressed into service as supposedly "free" lab networks.

Classify addresses at a glance

  • Check the type instead of guessing. The converter classifies every address against the IANA registries and names the range, plus the RFC for special-purpose ranges. That tells you instantly whether an address from a log is public, comes from your provider's CGN range 100.64.0.0/10, or is just a documentation example.
  • Find hidden IPv4 addresses. If a log shows an IPv6 address starting with 64:ff9b::, ::ffff:, 2002:, or 2001:0:, there's an IPv4 address inside. The converter shows it as Embedded IPv4, Embedded IPv4 (6to4), or, for Teredo, as Teredo server and Teredo client (public IPv4:port).
  • Don't build reverse DNS names by hand. A PTR record needs the address in reverse under in-addr.arpa, or for IPv6 all 32 nibbles in reverse under ip6.arpa. The Reverse DNS row in the converter gives you the name ready to copy — with IPv6 you'd almost certainly mistype it otherwise.
  • Zone IDs don't belong in other machines' configs. fe80::1%eth0 is only unique on the machine where eth0 exists. The converter accepts the notation and shows the zone separately; the address itself is link-local and isn't routed.

Combining with other JPKCom tools

  • From the IP to the domain. Once you've classified an address or planned a subnet, the next step often goes one level up — with the DNS, SSL, Redirect & URL Tools you check DNS records, certificates, and redirects for the domain behind it.
  • Placing your own location. The My IP tab shows your address and detected proxy headers; for the full picture of your connection and browser, the Server & Browser Info rounds it out.

More context: the overview for the big picture, the manual for every field, and the examples for the step-by-step calculations. You can try everything directly in the tool.