Skip to main content

A year ago I clarified my views on the #gemini protocol. I have since been told many times that I have misunderstood it. Sure.

daniel.haxx.se/blog/2023/05/28…

The Gemini protocol seen by this HTTP client person | daniel.haxx.se

daniel.haxx.se
#gemini
in reply to daniel:// stenberg://

Binary Large Octopus
Binary Large Octopus
• •
The strong recommendation for TOFU seems outright insane in this setting. Adding to the list of already stated problems comes the lack of revocation and certificate rotation methods.
in reply to daniel:// stenberg://

asyncmeow (pearl)
asyncmeow (pearl)
• •
"i have since bern told many times that i have misunderstood it" - do people think misunderstanding a spec that isn't written in a way to avoid ambiguity is a surprise? it's hard to properly understand a poorly written spec...
in reply to daniel:// stenberg://

Floating Point Error
Floating Point Error
• •

I read a post of some guy roasting Gemini the other day, in parallel to completing my Gemini setup.

Apparently that person was you 😅

in reply to Floating Point Error

daniel:// stenberg://
daniel:// stenberg://
• •
@aartaka I don't think "roasting" is a fair description. I think I provided honest and even quite constructive feedback.
@Floating Point Error
in reply to daniel:// stenberg://

Martin Rocket
Martin Rocket
• •
I read the post a while ago, and I liked it. I read the reactions, and I was disappointed how little the practical feedback was appreciated.
in reply to daniel:// stenberg://

Ed Davies
Ed Davies
• •
…which I find peculiar because a normal RFC 3986 URL is just a set of plain octets…


No, RFC 3986 is explicit that a URL is a sequence of characters. E.g., the first bullet point in 1.2.1:

A URI is a sequence of characters that is not always represented as a sequence of octets.


A URL on a advertising bill board is not a sequence (or “set”) of octets but it's still a URL.

in reply to daniel:// stenberg://

Ed Davies
Ed Davies
• •
Yes, of course everybody knows you're an authority on how URLs work, which is why an aside that misleading shouldn't be allowed to stand unremarked. No need to be sarcastic about it.
in reply to Ed Davies

daniel:// stenberg://
daniel:// stenberg://
• •

@edavies I did not quote the RFC and I did not use the RFC language in my post because I was talking about how URLs actually work. Not what RFC 3986 says.

URIs are in fact 7 bit ASCII for all practical purposes, so I can't think of any real-world case where a URI character is not an octet.

"URLs" and "IRIs" however...

@Ed Davies
in reply to daniel:// stenberg://

Ed Davies
Ed Davies
• •

Indeed.

But I still think the wording could be better to make that distinction clear so people won't be mislead. Just a prefix of “for practical purposes” would go a long way.

(Real-world exception to 7-bit ASCII octets: a few characters in SMS text messages where '[' and ']' (e.g. for an IPv6 address) are escaped. But all buried in the messaging apps so not much of a practical issue for anybody not programming those.)

BTW, I really liked the rest of your take on Gemini. I looked into it a bit quite a while ago and liked a lot, particularly the client certificate idea in principle, disliked some aspects of the markup language (lack of images in text documents) but was really uncomfortable about the scanty and mixed up specifications. Thank you for taking the time to articulate similar ideas in so much more detail than I had the patience for.

⇧