Hmm. I shoulda seen that coming ...
When mirroring an LJ entry to GJ and IJ, if the entry contains
an <lj ...> tag (e.g. <lj user="examplename">
=>
examplename), the IJ and GJ copies of the entry
will interpret the same tag as linking to a user of the same name
on those sites. And similarly, if I try to link to a GJ
member, the LJ copy of the entry will try to point to an LJ user
with that name. Oof.
Unless TPTB decide to extend the <lj user=""> tag to have syntax that can point to other LJ-codebase sites (or maybe other OpenID sites in general), and all three sites install that feature, I've got a problem. Maybe I should just forego the convenience of <lj user=""> and <lj comm=""> and use longhand HTML to produce the same effect instead (maybe stick a macro in my .exrc file to help) ... but I think that would break the "automatically fix all your old tags to point to someone's new name when they rename their journal" feature, wouldn't it?
Thoughts? Suggestions? Obvious workarounds I've overlooked?
I'm glad I noticed this before getting around to copying all of my old LJ entries over to IJ and GJ.
[Edit: a test of doing things longhand (though I might simplify this a bit from the copy/paste version here, in practive):
This is what the <lj user="dglenn"> tag expands to on the three sites, minus a <span> tag that must reference a CSS style with too many side effects; for simplicity on my end, I could leave off the little symbol that links to someone's profile (note that GJ uses the same icon as LJ) and just do a plain old href on their name ... would folks miss the "this is an LJ/GJ/IJ user I'm pointing to" indicator and the convenient profile link, or does it not matter much?Hmm. IJ doesn't link from the icon, and links the username to the profile instead of the recent-entries page -- odd. Bug?]
(no subject)
(no subject)
Testing
Re: Testing
I know you've solved this now, but just as one datum: when I see an LJ name in a post and am curious, it's the profile and not the recent entries I click on. I want to see the overview before diving into the journal, and, well, there are a lot of unreadably-formatted journals out there anyway.
Re: Testing
Going the other direction, I'd randomly gotten logged out from LJ at the time I tried to post a comment to IJ as my LJ self, so I got a "you need to be logged in" page -- I logged in in another tab, refreshed the first tab, and from there things worked the same as they had in the first experiment.
I hadn't thought to try doing the same thing with GJ, and hadn't heard that GJ doesn't support OpenID. :-( Sure enough, there's no OpenID button on the GJ comment form. That will definitely be a problem if it comes to folks who stay on LJ wanting to comment to folks who move to GJ. Ouch. And the impression that I've gotten from reading a lot of comments during the peak of the
Strikethrough2007discussion, is that a lot more of the people planning/considering jumping ship are leaning toward GJ than IJ.I just tried it the other way, posting here as my GJ-self, and it didn't work -- I got an error message about there not being a relevant authentication server registered. So anyone jumping ship to GJ won't be able to post non-anonymous comments to their still-on-LJ friends unless they keep an LJ account. (Presumably a free one ...) Then again, we'll still need accounts on the same sites as our friends to see friends-locked entries, regardless of the OpenID issue, unless there's some funky/useful interaction between OpenID and RSS. (Would that be safe, or a big security hole?)
I don't know whether, ultimately, that's good or bad for GJ. Will the lack of OpenID support impell LJ users whose friends have moved to GJ to follow them, or will it cause people who would have jumped to GJ to choose IJ or DeadJournal instead so that their still-on-LJ friends can comment easily? (Note: I haven't verified that DJ supports OpenID yet.) And if GJ, which AFAICT has neither ads nor paid accounts, suddenly swells to a huge size due to migration, will it be able to support the traffic? Though I worry about splitting comment discussions into three smaller fragments by mirroring, uncertainty about the unfamiliar sites (just because they are less familiar) made redundancy seem a sensible precaution.
First things first ... and first is sorting out whether and how far to trust LJ, and working on my tools to mitigate the pain if LJ does something amazingly stupid and breaks itself, or most of my friends migrate, or I'm pushed to punting LJ in protest. (I don't want to wave the "I'll leave" stick in LJ's face until I know I'm ready to follow through if they call my bluff. Though having LJ turn out to be trustworthy after all and staying here would, of course, still be the least painful / least annoying option.) Worrying about others' business models comes second, a luxury resulting from redundancy.
Re: Testing
Yah, more often than not, I do that too (and I share your pain regarding unreadably-formatted journals). I wasn't sure how many other people did the same thing.
Okay, I'm definitely keeping the profile link thingie. The other question is whether the pop-up box that has "Add friend / View profile / View journal" when you hover over the icon ... is useful, is mostly ignored, or is annoying as hell to most people. I'm leaning toward just not bothering to preserve that feature, both for the sake of simplicity and because I find it gets in my way most of the time. (If I've got the mouse on the icon, it's because I'm about to click it to view the profile; if I want to read the journal, I'll just click on the name, and I'm not going to add them as a friend without peeking at the profile and usually their recent entries first. But do most people like it?)
(no subject)
(no subject)
If I leave off the little link-to-profile symbol, it'll be faster to just type the HREFs by hand; if I do include that, I'll want to make a macro for vi -- I think that's right about where the hassle/convenience boundary is for me.
I'm guessing that the pop-up "add friend|view profile|view journal" box is code embedded in the style referenced in the <span> tag that the &lj;user> tags expand to; I'll probably just throw that out unless it turns out that a lot of other people find it especially useful.
(no subject)
(no subject)
Can I just blame this oversight on the general brain-friedness (http://dglenn.livejournal.com/947331.html) I complained about earlier?
(no subject)
the entries in XML and have stylesheets to whack it into all the target dialects.
(no subject)
If I were using (writing) a GUI client, it would make even more sense to have the front-end tool create XML as an intermediate product. But I've been writing my entries in HTML by hand in 'vi', and I'm comfortable with that, and I get the impression that (in general) XML is more useful as a language for two programs to use share data with each other, than as something humans type. Am I wrong about that? (Admittedly, my impression is gleaned partly by looking at MusicXML and comparing it to ABC and Lillypond.)
(no subject)
I do with it (probably similar to the things you do). The more I tweak and customize the
HTML, the better XML looks, as I can apply the tweaks automatically and uniformly by
simply reprocessing the XML originals (I edit mine with vi for the most part).
However, transforming XML with stylesheets can quickly get into deep magic if you
want to get creative (which I'll take as a given). Personally, I have a great deal of
fun making stylesheets do intricate things (e.g. stylesheets generating stylesheets,
stylesheets transforming XML into troff), but "normal" people try to avoid such things.
I even have stylesheets that take parameters from the command line, do I18N on the
fly, and call Java for various tasks.
(no subject)
(no subject)
I've still got a lot of work to do on tools. I need to improve the wrapper script to handle multi-word subjects correctly or hack the 'clive' source code to post to multiple sites simultaneously; I should add <lj user> tag expansiom to either 'clive' or my wrapper script as
So right now I'm only a fraction of the way there. A particularly visible fraction, I guess.
I'm also thinking that I should consolidate copies of all comments from all three sites into one pile in a completely different place for the convenience of anyone who wants to see all the comments without going to three different URLs, though I'm not sure whether a reader there who wants to add a new comment should just post it there or be redirected to LJ/GJ/IJ. (I'm leaning toward redirecting them and making the consolidation repository read-only, to avoid having to deal with identification/authentication issues myself -- and so that a reply to an LJ comment read there gets posted into the thread on LJ, etc.)
I'm not sure whether to set up RSS feeds between the three LJ-codebase sites or not. While all three accounts are active it seems like a pointless redundancy, but if one account should go *poof* for some reason, having already set up RSS feeds from the other two services on that one would be good.
Er ... so not only am I still building the tools, I'm still figuring out exactly how to approach the problem.
And I haven't even started to deal with how to handle SMS posts from my cell phone. Or voice posts, if I ever decide to use that feature. Maybe a 'cron' job that looks for un-mirrored changes to each journal and clones the new entries as it finds them (with a delay and recheck to make sure it didn't just happen to catch something still in the process of being posted to all three)? Could I pick an interval frequent enough to be useful but seldom enough to not be abusive to any of the three sites?