Subtle changes, bugs and feedback

I have no idea what any of it means but joining in with the cheerleading

Thanks.

And with that...

Turned out to be as simple as this: https://github.com/microcosm-cc/microweb/commit/99842e25c0f836090589c31a00677f576b911eee

Though it was subtle... it was called implicitly via comments.js as well as explicitly on the page. Sneaky JS developers.

Though it was subtle... it was called implicitly via comments.js as well as explicitly on the page. Sneaky JS developers

Yes! Yes! Yes!

Thank you.

Thank you.

No worries :)

Thanks indeed. Check your stats, I suspect the quoting function may make see an uptick in usage.

And with that ...

Quoting is still broken (#inb4tester)

But srs, glad that bit of it has been fixed

You think I have stats on this stuff?

Quoting is still broken

The failure to quote links?

I think I have a solution to that having looked at the code yesterday.

It is doing this:

    comments.prototype.getWindowSelectedText = function(){

      var selection     = window.getSelection();

      if (selection && typeof selection.toString == 'function') {
        return selection.toString().replace(/\n/g, "\n> ");
      }

      return false;
    };

Note the .toString() in there.

It would be better if it copied the HTML fragment.

Perhaps something along the lines of:

comments.prototype.getWindowSelectedHTML = function(){
  var selection = window.getSelection();
  
  if (selection && typeof selection.getRangeAt == 'function') {
    // Pull a balanced tree HTML fragment from the DOM for the given selection
    var range = selection.getRangeAt(0);
    var fragment = range.cloneContents();

    // We only want core HTML and link attributes, almost every other attribute
    // can be jettisoned
    var walk = function(node,func) {
        func(node);
        node = node.firstChild;
        while(node) {
            walk(node,func);
            node = node.nextSibling;
        }
    };
    var stripAttrs = function(node) {
      if (node.attributes) {
        var names = [];
        for (var i = 0; i < node.attributes.length; i++) {
          switch (node.attributes[i].name) {
            case "href":
              break;
            case "title":
              break;
            case "alt":
              break;
            default:
              name[names.length] = node.attributes[i].name;
          }
        }
        for (var i = 0; i < names.length; i++) {
          node.removeAttribute(names[i]);
        }
      }
    } 
    walk(fragment, stripAttrs);

    return fragment;
  }
  
  return false;
};

I wrote that blind... I've no idea whether it would work. I'm just guessing. I wish I had a JS developer around.

Oh. Sorry, no.

Was a /sarc comment about how replies worked much better in the old forum

That's OK.

I wrote a working version of that idea...

var getWindowSelectedHTML = function() {
  var selection = window.getSelection();
  
  if (selection && typeof selection.getRangeAt == 'function') {
    // Pull a balanced tree HTML fragment from the DOM for the given selection
    var range = selection.getRangeAt(0);
    var fragment = range.cloneContents();

    // We only want core HTML and link attributes, almost every other attribute
    // can be jettisoned
    var walk = function(node,f) {
      f(node);
      var child = node.firstChild;
      while(child) {
        walk(child,f);
        child = (child.nextSibling) ? child.nextSibling : false;
      }
      return node;
    };
    var stripAttrs = function(node) {
      if (typeof node.hasAttributes == 'function' && node.hasAttributes()) {
        var names = [];
        for (var i = 0; i < node.attributes.length; i++) {
          switch (node.attributes[i].name) {
            case "href":
              break;
            case "title":
              break;
            case "alt":
              break;
            default:
              names[names.length] = node.attributes[i].name;
          }
        }
        for (var i = 0; i < names.length; i++) {
          node.removeAttribute(names[i]);
        }
      }
    }
    fragment = walk(fragment, stripAttrs);

    // Now we have a clean fragment we need to get the HTML as text but a
    // DocumentFragment does not have an innerHTML so we need to attach it to
    // the DOM and ask that for the innerHTML.
    var scratch = document.getElementById('scratchpad');
    if (scratch) {
      while (scratch.firstChild) {
        scratch.removeChild(scratch.firstChild);
      }
    } else {
      scratch = document.createElement("div");
      scratch.id = "scratchpad";
      scratch.style = "display: none;";
      scratch.hidden = true;
      document.body.appendChild(scratch)
    }
    scratch.appendChild(fragment);
    return scratch.innerHTML;
  }
  
  return false;
};

I'm not sure whether I'll use it, because HTML scares people. But the idea is nice.

It returns the HTML of the selected text, and so would preserve links, formatting, etc... whilst ensuring that the HTML is not broken.

The only downside is that in the text box that people type their reply, the HTML would show.

Could it be implemented as an opt in option for users?
Thanks for keeping on hammering away at these issues though :)

Users have options?

No, I guess not. Although there is some variation between being an admin/user in terms of what you can do?

But I expect this is another case where it looks like it might be very easy from the front end, but actually isn't at all

Although there is some variation between being an admin/user in terms of what you can do?

Nope.

Very equal.

We see the same thing, the only difference is a few permissions grant me the edit functionality on other people's posts.

There is very little I can do that you cannot. I wanted it that way, to avoid creating forums that had crazy dictators on them. I have a bit of control, but not some crazy admin control panel.

Also... I found that good defaults on forums were far better for 99% of forums than a panel full of options, any of which could negatively change the site if they were set without the admin understanding the option.

In essence: Admins are just users with a few more permissions.

Yep that all makes sense :)

Not sure how and if it's a consistent thing, but I think on mobile browsers I sometimes end up with the

enter code here

section pre-filled in the comment reply

Fat fingers?

What browser and OS?

You think I have stats on this stuff?

Implements new feature
Ignores metrics

Hey.....You could work on my product team!

I find it's really sensitive on mobile, even as you're scrolling it will think you've hit a button. The list button is in the middle so that's what I seem to get most of the time. Only seems to happen on this site.

android 6, chrome, xperia z3

@Velocio

how come I sometimes get a "secure connection not establishable" (bad translation from dutch) on my mobile. it seems to pop up at random and then I cant access lfgss for a while. it then vanishes without reason. what am I doing wrong?

Ooh, that's good feedback.

I'll see whether we are doing anything that affects the typical response times to clicks, swipes, etc.

We use HSTS now, which dictates that every page must be served over SSL.

Sometimes, and especially on mobile networks, the SSL handshake may be interrupted by bad signal. In that moment the browser has nothing... so the only thing it can do is to tell you this and let you retry.

We use HSTS now,

I still use HTFU

Ah, but HSTS is HTTP Strict Transport Security.

And in things like Chrome, we actually ship as a line hard-coded into Chrome to only ever access this site via HTTPS:
https://cs.chromium.org/chromium/src/net/http/transport_security_state_static.json?l=13230

Likely to be caused by using a mousedown / mouseup / touchend etc event on mobile to speed things up? The basic 'click' event on a phone introduces delay as it tries to disambiguate a genuine click from a scroll event. That delay might be exactly what you do want in this instance.