Reinforcement Learning and
Artificial
Intelligence (RLAI)
Suggestions Archive
The ambition of this
page is to keep an archive of the suggestions page.
The suggestions below have already been taken into account and hence
have been removed from the suggestions page.
Automatically generated back link on missing pages
>>I think it
would be great if new pages that
were added had a javascript "back" link built in to them. Or, a hard
link back to the main page. Something like that.
Otherwise its hard to find your way back if people don't manually make
a link.
Also,
while I'm at it - name the main page index.html so the url can be
shorter (so we only need to use the folder not the entire path)
[Rich: No, No, forget about file names! That way lies the old way
of thinking!]
[Anna: Definitely long paths are a problem, though. Magic users can set
up .htaccess files to forward appropriately.]
I think we should consider this Done.
>>The missing.shtml should be
altered,
first
to provide a link back to the previous page (from whence the missing
page was accessed) as suggested above. The link should appear as
back to /callingpage.html
except in a smaller font and flush right. DONE - not small - where
should it go? - check it out nonexistent
page A: that's great. put the link
just at the bottom of the ambition statement area, as suggested rewardhypothesis.html. -rich
Second, in the line "The ambition of this web page is to..." remove the
word "web". DONE
Third, in the initial mouseover text, instead of "[Anonymous, Jan 1
2004]", put just the current date.
DONE
Good work. -rich >>Oops. New/missing pages appear with the
automatic back link to the page that called them, but when you actually
go to create the page by editing it, the back link disappears. Thus it
doesn't work at all at present.
>>No, back links aren't implemented. The simple way I tried to do
it
doesn't work (because Mozilla's grabbing of the page for editing is an
isolated get request, we can't use the referrering page to construct
the back link and have to either use Apache's log file or start setting
cookies). So whatever is done for back links, it has to be done from
scratch.
>>I discovered that the error document
(missing.shtml) can be a cgi script (missing.py for example) and so I
thought I'd be able to do the back link thing that way. cgi scripts
have an environment variable HTTP_REFERER that refers to the page that
led to the current one. So, that would seem to work, but if you just
type a missing page into your browser the HTTP_REFERER variable will be
empty. Is this ok? Can we assume that all new pages will have a link to
them before they are created? Anna, were you using Javascript when you
implemented back links?
>>Using the HTTP_REFERER variable of a cgi script is exactly the
same as
using it in an SSL file, because the cgi programs just get the value
from apache. No, it isn't going to work. If you click on a link to a
missing page, you will see the back link. Then when you hit command-E
to edit it, the back link will disappear because when you hit command-E
Mozilla makes a totally new get request - exactly equivalent to typing
the URL into the browser. So HTTP_REFERER will not be set because you
weren't refered to the page.
But there's no reason missing.shtml needs to be an SSL file - a python
cgi file is fine. I'm not sure which is the active one, check the
httpd.conf file and delete whichever one it doesn't refer to.
>>
If we want to link to the last page accessed by that user, then even if
you type in a new URL directly, it will make up a link for you (as
opposed to having to make the link then click on it). In that case the
best approach is probably to have the missing page program look up in
the Apache logs to find out the last file viewed by the user at that IP
address - if it can. I think it can look in the logs (we might have
permissions problems). That's a bit nicer than a cookie because some
people are gun-shy about information being stored on *their* computers,
and if they have really tight security settings they'll get warnings on
every page, which would be annoying.
>>Ok! Back links are done! All newly created
pages automatically have a link to the last page the user viewed on
rlai. And when they select Edit Page the link stays. I used the apache
logs to search for the last page from that ip address that was not the
current page. Let me know if there are any problems with it.
Extension notification to include text of the extension
>>I
think we want to change the notification
of extensions to include the text of the extension. DONE by somebody. good
work.
>>We
need a "diff" feature, or something,
because I get (or send) a notification that something has changed, and
there's no clear way of knowning what all is new. If it's an extension,
that's easy to figure out, but if it's comments added in the body,
that's hard.
So maybe I shouldn't be adding
comments to the
body. But without any kind of threading of the discussion, it gets
really cluttered really quickly.
Do we expect people to do (via
HTML) their own threading? Their own highlighting?
A: I think it is not unreasonable to
ask those who edit the page body, and
who want to bring their changes to the attention of subscribers, to
include a substantial description of their changes in the notification
message. Remember that message can contain great swaths of the
modified text. And anyway i can't think of a better way.
-rich
Displaying list of subscribers to a page
>>I am thinking we should make the list of
subscribers visible to users. This would help in a lot of ways, and can
be done in a way that prevents potential downsides.
The benefits
are that 1) people can tell if they are already subscribed, 2) people
can tell if anyone else has subscribed (and thus might respond to their
note), 3) people can tell how much interest there is in their page, and
generally 4) there will be some balance between the invisibility of the
lurkers viewing and the authors out there with their ideas exposed.
The
potential downside is that people's email addresses become visible and
might be picked up by machines or harassers. However, this can be dealt
with. I suggest just subtly corrupting the addresses, say by changing
one letter at random in the username and domain name. That should be
enough to stop machines while still serving the other purposes.
[Anna: Yeeees, one letter would solve the bot problem, but not the
people harvesters. What about only showing, say, the first 4 letters of
the username and the first 4 letters of the domain name? Or some
obscure configuration. And we better warn people about that when they
subscribe, or we get in trouble.
Or, we could simply, at the bottom of each page, display the number of
people subscribed. We still need a simple "check if I'm subscribed"
mechanism, but I think that can be folded into the subscription pop-up.
Knowing that there are people subscribed is really important.
For balancing the invisibility of lurkers and the exposure of authors,
we could allow the author of a page to view the subscription list.]
A decision was made on this at
the October 20 meeting - we will enable the display of subscribers with
just their usernames with no domain names, or alternately with a
name
provided by the subscriber (could be anonymous). The subscriber
list will appear on clicking a "subscribers" link in the bottom
material.
>>I really want to be able to see the
subscribers! -rich
The list is now visible. I have to yet add the option to let people
have their names displayed instead of their email logins.
Setting title for a new page(missing page)
>>Newly created pages should automatically set
the title to be whatever is the first line enclosed by h1 tags, if it
is the default title.
>>
If a new page is created and the title is not set, then it is taken
from the first h1 field. There is a problem with this though, because
the missing page doesn't use h1 for it's title heading, so you have to
explicity switch to using h1.
>>missing.shtml doesn't seem to work properly in Internet
Explorer 6.0
under Windows XP. Non-existant pages come up with a generic error "This
page cannot be found". Haven't tried Internet Explorer under OS X.
>>That's weird about missing not working in
Internet Explorer - the first thing that comes to mind is that IE might
not properly respond to the 401 code (or maybe it's the 403 code, and
that's why it's not properly responding). The best test would be to see
if it displays the custom missing page of *any* site, and if it
doesn't, it's a bad program.
>>Ok, so I put the ErrorDocument directive in
the .htaccess file and now missing.shtml works! Apparently IE doesn't
like it in httpd.conf only... Are there any reasons it shouldn't be in
.htaccess? Also, there are 2 versions of missing.shtml:
/cgi-bin/missing.shtml and /openpageinfrastructure/missing.shtml. Which
is the right one?
>>No, no reason why it shouldn't be in the
.htaccess file. I was moving things into the httpd.conf file because it
is faster to not use .htaccess files, but we're going to need them for
passwords, so putting other stuff in there doesn't matter.
Visitor count on pages
>>Something about page visit counts would be
good. Maybe a bottom link to see the count. To help page maintainers
know if they are reaching anybody. -rich
>>Ok, so the viewcount on every page is up.
It's at the bottom right of every page. Also, there's a link to the
Archive at the bottom of every page. Overwrite protection is also done.
If you start editing a page in Mozilla and someone else changes it in
the mean time then you won't be able to publish.
>>The visitor count at the bottom of each page
has been changed to just a number which is a link to a page with more
information. Currently it's counting visitors, unique visitors,
extensions and edits. I imagine in the future it will become more
detailed and more explanatory, but the basic idea is there.
Email notifications
>>The subject line of the email notification should begin with
the meaningful part of the file name, e.g., instead of
Change to open web page
/openpageinfrastructure/suggestions.html
we should get
openpageinfrastructure/suggestions.html has been changed.
[For OpenPages 2.0, going one step further, we should use in the
subject line
something more meaningful than file
names. We should use the title/header of the page.]
>>The emails that OpenPages sends should be formatted better.
With a
proper from and to field. And the subject line has been shortened.
Also, these changes in the emails are now in extension notifications
AND general notifications (when you click on notify).
That's all for now. Formatting the cvs diff in the archive is proving
to be quite challenging.
Archiving for open pages
>>Archiving - Finished! Anything else needed? Possible additions:
a link to the different versions of a page on each page, a way to
automatically revert to an old version (maybe not a good idea), a nicer
looking diff display. [ Mark Lee, Mon Jan 24 14:28:09 2005]
>>Ok, so basic archiving is up. Openpages is
being archived in a CVS repository (Subversion was causing many
problems). The archive can be viewed online using ViewCVS (I know it
looks kinda ugly right now). The address for that is OpenpagesArchive/viewcvs.cgi/www/
I don't think it's possible to create some kind of archive that is
truly impossible to destroy. Pretty much if it can be created then it
can be destroyed. But, we can make it so that only administrators have
that potential, and so as long as they do things correctly the archive
should be stable.
