Recognizable Endpoint Ids in case of future success

[TotalKrill]

I’ve been thinking about how Iroh endpoint IDs are represented, and I kind of wish they had a more recognizable serialized form — something you can identify at a glance, like other addressing schemes.

For example, even though IPv4 is just 32 bits and IPv6 is 128 bits, you can instantly tell them apart when you see them. The same goes for MAC addresses, UUIDs, and other well-established identifiers — they’ve developed a familiar “look” that makes them easy to spot and reason about.

My utopian dream is that 5–10 years from now, when I stumble upon an old config file or API response, I’ll immediately recognize an Iroh address for what it is. I think the Display and FromStr implementations in Iroh will be key to shaping that kind of long-term recognizability.

To get there, I’d love if the serialized form of an Iroh address could aim for a few practical traits:

• URL-safe (no +, /, =, or reserved symbols)
• Case-insensitive (ABCD = abcd)
• Easy to parse
• Reasonably short (ideally <64 characters, so it’s DNS-friendly)
• Visually distinctive enough that sysadmins and developers start to recognize it on sight

Hex encoding already satisfies the first few points, but it’s not particularly compact or recognizable. A Crockford Base32
representation would check nearly all the boxes, and adding a short prefix could make it even more distinctive in the long run.

[TotalKrill]

After thinking about it a bit more, I’d suggest using Crockford Base32 for serialization, with a single - separator to split the short and long forms.

Example:

short: 6zirk5
long: 6zirk5-mha2nh5kql2i6hriadkiaujgnwlrqb5n2as5itzj5zgxea

That way, the short form can stand alone when scanning logs or reasoning about code, and the long form is visually distinct from any other addressing scheme I found.

[flub]

I do like this suggestion. Definitely also having the short-form clearly visible in the long form.

[flub]

I would also really like the check symbol of the Crockford Base32 to be part of this. I don' think you've used that in the example above.

The short form is something you came up with yourself or is written down somewhere with all the tradeoffs?

[TotalKrill]

The short form is just something I randomly chose. I generated that address from iroh v0.29 that actually used base32 formatting, and it was visually distinct to me at least. But I found the short form to be too long, and since its unly usable as debugging / reasoning, I thought 5 characters were a good choice.

The check symbol I chose to disregard, because it makes the address not url-safe

[flub]

FWIW in iroh's log we often use EndpointId::fmt_short uses 5 bytes.

[TotalKrill]

Makes sense, ill just use the same, made a PR for it #3609