Subtle changes, bugs and feedback

Vogon. (Or Bogan, I suppose. Which I imagine to be equally unpleasant)

But due to all edits being ninja, non-timestamped or visible you can correct your mistake and my comment will look like I'm seeing things. Any plan to implement in the near-future?

all edits being ninja, non-timestamped or visible

We store that info... I shall add it to the list of small things to do. Todo: Make visible when an item has been edited (if over 15 minutes).

I'm also going to tidy up the cluster-fuck that is "Actions"... which now looks like a mess of forms.

I've looked into the reported bug, and it isn't a bug.

At the time that you have taken the screenshot the "Following" page is loading, but hasn't finished loading. The progress bar shows that your phone is barely 10% of the way through, basically waiting for a connection.

In effect... you're not viewing the menu as if you were on the Following page, you are still viewing the entirety of the older page that you were on and that is the Forums page.

When the page does complete loading, then check the burger menu and you will see that the Following tab was highlighted.

Were there plans to introduce 'reply to PM by replying to the notification email' feature, or have I imagined this?

Yes, but we haven't got to the email stuff in general yet.

The gist is, we create a message ID with every notification, and if you reply to an email we pluck that back out and know what you are replying to and then process it as you would expect.

Additionally, we hope to support being able to email someone by their username and site, the equivalent of Velocio.lfgss@microcosm.app would check who the sender email address is, and if it's a known profile of the same site (LFGSS in this case), then it would create a PM shared across an audience of whomever else was CC'd.

We chose the right provider for this... mailgun. They offer an inbound email API that allows us to do this.

But it has a cost associated to it, and requires testing, and hasn't been our recent focus.

Ok, that's not how I thought it worked before - I seemed to remember it moving to show the active choice (like a button press or the desktop LFGSS menu).
I'm not sure of the point of seeing where you already are in the burger, other than it matches up visually with the non-menu part of the desktop site...

Any chance you could fix whatever it is that's stopping people from being able to @ mention me properly?

Either that or let me have @dan as my username?

@dan ?

Ah yes, I remember.

Teenslain can't be @ as well.

@_dan_ works if you insert a reverse slash after the @.

Good to know - but seeing as I don't @ myself...

Yeah it's a bug.

The "dan" username is reserved by an invester, @hamster has it. He may be willing to let go of it, but perhaps not.

And the bug can be summed up as this: The mentions are processed after Markdown is processed, which means that the Markdown in a username is processed before we look up usernames.

To process the mentions before the Markdown is processed (the bug fix) creates a different bug... namely that processing it in this order means that we do not know which parts of a comment will be turned into HTML block elements that should prevent mentions from being processed, such as <code> elements which are quoting the raw strings.

And it turned out, that applying the bug fix surfaced that new bug almost immediately on the sites that share code snippets such as the Espruino forum.

So right now... the workaround is not to put Markdown characters in your username.

But what should be the fix is that the Markdown library should apply a two pass processing, the first pass being to discover and create all blocks (including blockquote, table, code, etc), and then to allow custom code to be applied (process mentions) before finally doing the second pass to process inline Markdown.

Unfortunately the existing Markdown library is a very long way from doing that... they do everything in one pass and without an abstract syntax tree. It's not many steps removed from just running regular expressions.

Which means we either:
1) Run regexps ourselves, pre-Markdown processing
2) Run HTML aware processing post-Markdown processing

The first creates bugs all over the place.
The second creates a bug with usernames that contain Markdown.

Considering the impact of the second scenario is far less than the impact of the first, and that there is a work around (don't put Markdown characters in your username), that is the scenario I chose.

The characters to avoid in your username include *_=\ and then anything that isn't a display character, such as HTML entities that are not rendered like ­.

The issue may appear if you only have a single Markdown character in your username, but definitely will appear if you have a pair of the same character.

Why not call yourself @.dan. or something similar?

@(dan) ?

He's been taken, @)dan( ?

Or @5thDan ... then you can be Godan and a black belt in all things Karate.

danmit ?

@dandandandandandandandandandandandandandandandandandandandandandan

@_dan_pleaseputabackslashbetweenmynameandtheatsigntomakemynotificationswork?

Change your name to "dan " (without the quotes, note the space) as you've posted more than the reserved dan your username will be at the top of the list when people @ and it will look the same to them.

I'm getting:

Secure connection: fatal error (1066) https://www.lfgss.com/updates Unable to verify the Web site's identity (OCSP error). The response from the online certificate validation (OCSP) server was too old.

on Opera/9.80 (Windows NT 6.2; Win64; x64) Presto/2.12.388 Version/12.17 only, everything is fine on other browsers. Any idea why?

Any idea why?

Opera is crap and you are a masochist to continue using it?

Yup, seems to be the case: http://forums.opera.com/discussion/1773932/broken-websites-twitter-tumblr/p1

Restarting your browser may resolve it, disabling security features may resolve it, changing your browser will resolve it.

For fun I checked the certificate, re-read all the details, verified that it's valid and well-cached, checked the revocation status, confirmed that it has a non-SNI fallback, and re-verified the availability of the OCSP stapling and that the server in question is up. Everything is good... it's Opera.

Though in your favour the senior engineer at Opera did say it's only a temporal issue... if Opera borks the communication it only saves the dud status for 24 hours.

So you could try tomorrow.

So you could try tomorrow.

Being off LFGSS for a day is a much smaller sacrifice than putting up with the defects of "better" browsers, and I include later versions of Opera in those scare quotes. Having switched to Firefox to post, I'm reminded of how horrible LFGSS is on a big screen without css hacking, and I cba to hack css on a second browser.