majestrate / srndv2

some random news daemon (version 2)

srndv2

building:

go get -u -v github.com/majestrate/srndv2/cmd/nntpchan

srndv2's Issues

Add daemon mode and init scripts

This is long over due.

UTF-8 and Subject: - header of article. NNTP

Here is RFC 3977:

o Header values SHOULD use US-ASCII or an encoding based on it, such
as RFC 2047 [RFC2047], until such time as another approach has
been standardised. At present, 8-bit encodings (including UTF-8)
SHOULD NOT be used because they are likely to cause
interoperability problems

UTF-8 SHOULD NOT be used without " =?utf-8?q?this is some text?=
But, This specification extends NNTP from US-ASCII to UTF-8.

Should we use =?utf-8?q??=

Recache the board model when a new thread is created

When a new thread is created it potentially creates a new page, but currently this results in this error:
board page should not exist
because the new page isn't added to the board model.

auto reload daemon on runtime after recompiling

Whenever I update i have to stop the daemon, git pull and recompile, the process is unacceptably too long.

I think we need a way to reload the daemon live via signal, i.e. SIGUSR1

Can't touch the running binary executable so this would require a wrapper program that loads/unloads a shared library module ? Some other way?

>>fffffffff ImageBoard links for NNTP transfer.

Hello. I am working at diploma about your nntpchan. It is Java implementation.

We have problem with links in messages, like this:
>>fc346a235
We need NNTP format for links => convertation from your local
articles:message_id_hash ... to ... global message-id of NNTP.
When you feed(peering) with NNTP.

I will public my full work when it will be completed. See you.

Rebase template generation

right now I use a global singleton instance for generating templates which is a piss poor implementation.

there are right now, models, templates, cache.

models are in memory instances that hold state but also can render themselves

templates are like views, they render models by wrapping a model's rendering

cache is just a cache

ideally I'd like to use a classical MVC architecture but this would require reworking most of the frontend code as it is now.

mime type filtering

this has been put off for waaaay too long

NNTP attachment Content-Disposition header.

You have broke something in new commits
with Content-Disposition recognition.
I have such:
Content-Disposition: attachment; filename="name.png"

You can't display thumbnail.
Look at test board.

Rate limits

When creating 100s posts at once, I received:

{"error":"pq: sorry, too many clients already"} && {"error":"pq: remaining connection slots are reserved for non-replication superuser connections"}

rate limit is fine but needs to be documented. Rate-limiting to one every 100ms seemed to get around this.

Web API Problems

through nntp, I see this does exist
http://nsfl.tk/api/header?name=X-Mongo-ID&value=57260140c984bafa4cd5499e
but web api is returning []

Do ping pong!

Send "500\n" or "говно\n" to maintain stream connection once in 1-2 min. Make it neatly. Do not send it in the middle of another transfer.

[nntp] periodically check all peers for new posts

This is especially helpful when the post rate is really low with frequent netsplits.

Implement IP Range Bans for redis backend

introduced redis backend in 1301dad but lacks range ban capability.

expire orphaned articles

when doing sync, sometimes the root post is expired before we can get it so we have a bunch of replies with no root post.

ideally if we can't resolve these root posts the orphaned replies should be expired.

Potential rounding issues

In the postgres driver there are several instances of
page / int64(perpage)
which are converted to int again. This may lead to problems with floating point precision.
This is especially true for
( count / 10 ) + 1
Let's say you have exactly 10 threads here, so with this code you may end up with the unexpected result of 2 pages.

The correct way to handle this, would be to use math.Ceil

implement better captcha

existing captcha is probably too weak so we'd need to implement a better one.

use /usr/bin/identify to store information about thumbnail size

Network latency and slow nodes tend to cause pages to jump since loading an image of unknown height and width causes reflow.

/usr/bin/identify is an ImageMagick command that returns the name, file type, width x height, [some weird string that has the width and height in it], bit depth (?), color space (?), file size, [some thing I don't understand], [some thing I don't understand].

$ /usr/bin/identify nntpchan.png
nntpchan.png PNG 2508x2043 2508x2043+0+0 8-bit sRGB 282KB 0.000u 0:00.000

If this information can be provided to templates, it will help improve usability.

Limit the number of characters for the comment field

Posting this for example locks the whole thing up for a few seconds.