💾 Archived View for gemi.dev › gemini-mailing-list › 000226.gmi captured on 2024-08-25 at 10:15:29. Gemini links have been rewritten to link to archived content
⬅️ Previous capture (2023-12-28)
-=-=-=-=-=-=-
Hi everyone
Just to toss a couple of things into the debate, for those interested in
a minimal application capability for Gemini, but don't want all the
bells, whistles and complexity of what the web became (cue scary music)
I did a bit of digging on the background history of HTTP. Maybe you are
all familiar with this, but personally I found it quite illuminating on
our conversation. If we are positioning Gemini between Gopher and the
Web, we can decide where is the sweet spot along the spectrum.
https://medium.com/platform-engineer/evolution-of-http-69cfe6531ba0
https://developer.mozilla.org/en-US/docs/Web/HTTP/Basics_of_HTTP/Evolution_of_HTTP
https://hpbn.co/brief-history-of-http/
HTTP 1.1 was the first fully standardised version, having an extensible
structure, and many possible headers, cookies etc. However before that
there were two pre-standardisation version that are quite interesting to
look at.
== HTTP 0.9 ==
This was the very first version designed by Berners Lee. All it had was
GET (sound familiar?)
Modulo the certificates/TLS stuff in Gemini, it is more or less what
Gemini is right now.
If you just want idempotent resources and nothing else, this is all you
need and is what we have in Gemini at the moment.
== HTTP 1.0 ==
This introduced two other methods HEAD (a GET with no response body) and
POST which is for sending content to the server. It was still one
connection per resource. It defined URL format, charsets and mime types
and the idea of a header format and some standard headers
(Content-Length, Date, Others).
It also defined a mechanism to create new arbitrary headers (we can get
rid of that)
== HTTP 1.1 ==
The first fully standardised version that really took off and had a
million different options including persistent connections, cookies and
so on. This is too much for our needs and the extensibility opened the
door to all the sugary complexity we are trying to avoid.
_________
Looking at HTTP 1.0, it seems to mirror the basic needs for Gemini to be
able to be used as a platform without the bloat that became 1.1, 2.0
(and in the future 3.0/QUIC is in discussion)
My sense is we can find the sweet spot at about "HTTP 0.95" which is
enough to support application development without the cruft of arbitrary
headers and extensibility.
Here is the RFC for HTTP 1.0:
https://tools.ietf.org/html/rfc1945
quoting from the RFC about POST:
The POST method is used to request that the destination server accept
the entity enclosed in the request as a new subordinate of the
resource identified by the Request-URI in the Request-Line. POST is
designed to allow a uniform method to cover the following functions:
o Annotation of existing resources;
o Posting a message to a bulletin board, newsgroup, mailing list,
or similar group of articles;
o Providing a block of data, such as the result of submitting a
form, to a data-handling process;
o Extending a database through an append operation.
The actual function performed by the POST method is determined by the
server and is usually dependent on the Request-URI. The posted entity
is subordinate to that URI in the same way that a file is subordinate
to a directory containing it, a news article is subordinate to a
newsgroup to which it is posted, or a record is subordinate to a
database.
<continues>
A valid Content-Length is required on all HTTP/1.0 POST requests. An
HTTP/1.0 server should respond with a 400 (bad request) message if it
cannot determine the length of the request message's content.
Applications must not cache responses to a POST request because the
application has no way of knowing that the server would return an
equivalent response on some future request.
This seems to well describe the needs of building non-idempotent
applications in Gemini.
As per previous emails we dont need the full vocabulary of the other
HTTP verbs such as DELETE/PUT etc
and we can also still dispense with HEAD (a body-less GET) as adding not
much (and was not much used).
Clients can implement a simplistic form of HEAD by abandoning a
connection after the header line is seen.
So what I think we need is something around HTTP 0.95 (it never existed)
which is about half way between HTTP 0.9 and 1.0.
GET (gemini right now)
POST (for non-idempotent interactions)
Content-Length indicator (to make POST work)
no other headers, so not extensible, so not as far as HTTP 1.0, but
allows application development.
I'll leave you in peace now :-)
- Luke
On Tue, Jun 16, 2020 at 08:21:51PM +0100, Luke Emmet wrote: > So what I think we need is something around HTTP 0.95 (it never existed) > which is about half way between HTTP 0.9 and 1.0. > > GET (gemini right now) > POST (for non-idempotent interactions) > Content-Length indicator (to make POST work) > > no other headers, so not extensible, so not as far as HTTP 1.0, but allows > application development. This sounds really nice (honestly!), but the big problem for me was always: how does the client tell the server if it's doing a GET or a POST without the protocol including some kind of marker for this, which could then potentially get extended ad infinitum by people defining their own equivalents to HEAD and PUT and etc, etc? It's the usual theme: if you have *one* of something, it can be implicit, and as such nobody can try alternative values. As soon as you have two or more, you need to tell the server about it, and then people can try telling it different things, and before long there is more than just two... Cheers, Solderpunk
On 16-Jun-2020 21:00, solderpunk wrote: > On Tue, Jun 16, 2020 at 08:21:51PM +0100, Luke Emmet wrote: > >> So what I think we need is something around HTTP 0.95 (it never existed) >> which is about half way between HTTP 0.9 and 1.0. >> >> GET (gemini right now) >> POST (for non-idempotent interactions) >> Content-Length indicator (to make POST work) >> >> no other headers, so not extensible, so not as far as HTTP 1.0, but allows >> application development. > This sounds really nice (honestly!), but the big problem for me was > always: how does the client tell the server if it's doing a GET or a > POST without the protocol including some kind of marker for this, which > could then potentially get extended ad infinitum by people defining > their own equivalents to HEAD and PUT and etc, etc? It's the usual > theme: if you have *one* of something, it can be implicit, and as such > nobody can try alternative values. As soon as you have two or more, you > need to tell the server about it, and then people can try telling it > different things, and before long there is more than just two... One really simplistic solution is we just have another scheme name (foo:// gemini+foo://), or just bolt it down in the spec (like mime and charset is). Gemini doesn't have to specify its really a GET, because it only does GET. BTW I think there is lots of juicy ideas in your new blog post - still digesting it. Best wishes - Luke
---