Contents:
For fixed issues, please see the most recent entries in the History section.
Significant events and updates in the history of the Mobile Access Portal.
Talking to Orange in bath; research for a new phone.
At that point, I had an existing Orange phone dating from 2000, with a dead battery and no replacement batteries still available, and no fancy features, no camera, etc.
Colour screens as standard, remarked the salesperson. I was impressed with this, having previously only been able to make phone calls in black and white...
Brought shiny new phone at Orange shop in bath, with security strap from FoneSolutions.
Started the process of opening a business account at Lloyds, relevant because this was where the business was at this point.
Discovered that on my mobile, I could alter the links in the shortcut menu on the main screen, by editing the first three bookmarks. Put facebook, google, and flickr in there, and moved Orange's click-crash suggestions elsewhere.
This sort of ergonomic need is in the same theme as the reasons driving the development of the Portal.
Wrote the core of a "basic" filter app, to strip the junk out of saved pages like flickr and Wikipedia. I've called this !Defluff.
'Dashed off a brief for the idea:
Single input field with multiple submit buttons, for quickest possible response: Google (default) flickr text/tags/recent, BBC, Wikipedia, InterNIC Nominet.
If input field blank, go to that site's homepage instead.
Wrote various string/search/replace/regexp routines for !Defluff, and knitted them together with a parser I previously wrote for a logic circuit emulator.
Wrote and bunged online the first version: Best accessed from a smartphone.
At this stage, purely one documentation page, portal homepage, and the form handler.
Several alterations and expansions to the portal. Tested the new php on the offline server before letting it loose on the world, and was glad I did!
Portal now in Beta!
Found correct Yahoo email link. Not easy, because I don't access Yahoo email via mobile web, but by a bundled 'n' locked mobile app instead. Looked at adding bebo.
Testing the buttons:
Gave mobile search portal working title of "MAP".
Lots of offline work.
Mobile Access Portal now mostly-debugged.
Added default font face as "System", in FONT element, because not clear how to do this in CSS. Online documentation is now Valid HTML, apart from the use of the FONT element.
Some problems with Sky News search and Hotmail login, due to slow connection from those Servers at current location.
Blank-usage button tests.
Social networking searches okay, apart from facebook group search.
At this stage gmail and Yahoo were nearing the end of tests, and plain links also still contained News searches, Buzz, and IP lookup.
"This IP" works, but produces too "heavy" a page. To switch to a self-provided service later.
Idea for clock at bottom: "Later than 00:00", because the clock wouldn't be able to update.
First intensive offline working for a while, due to being clobbered by a virus for the last week. Virus interrupted thinking for the beta feedback facility, or rather it's anti-spam techniques.
Adapted HTML syntax colourer from php.net documentation.
'Wanted to get on with the last Portal updates, and do things like move the "Why?" link, so it doesn't get first tab order, instead of the search query field.
'Aim later to write the reporting part of a specialist search engine, partly to assist a third-party service provider.
Also want to add bits to the Portal, including replacing the InterNIC and Nominet options with a [universal] whois. Would like to get mobile Buzz and hunch searches available, too.
Got the Portal updated (offline), taken about as far as I can go without online testing for now. Got the source viewer linked into the Portal, expanded the documentation, added test adverts, made the source viewer work with (offline-simulated) remote documents.
Quick-ish tests in cold, noisy, but very handily-placed intercafe, and 10 further hours scripting away in the office.
BBC's copy of Betsie won't access non-BBC sites, so downloaded a copy ready to use on our own servers, knocking out the checks for a "safe list" along the way.
Back into the office for another stint, passing free internet access on the way.
Source viewer online, and work's perfectly, first time!
HTML & CSS Validation.
Started plugging in hunch.
Tested to see if Perl ability (for Betsie) still there. (It was.) Betsie working roughly on portal host.
HEAD request tests.
Somehow managed to transpose hunch and php lookup code during offline debugging, so they carried out each other's functions. Corrected this.
Checked internal links [anchors] of documentation FAQ, thus ensuring the contents "index" was reliable.
Lots of offline updates.
All stable portions now XHTML-Compliant, including their output.
A favicon designed and added: Quickly sketched, bunged in a !Draw (vector design), grabbed as an image, fiddled about with for colour, scaled/anti-aliased to icon size, converted to .ico format by GIMP2.
Early HEAD request analysis tool working.
Added the new CSS I nabbed from facebook to a test page with new-style buttons, then tested it on the mobile: No effect! Curses - back to the drawing board.
Neither php nor the Validator liked the XML declarations; the <?xml .....?> bits on the first lines of XHTML files. Fixed this by substituting them with a line of php code, escaping all the dubious characters.
Fixed Trainwest (beta advertiser) link (they were using frames).
Redirected Buzzzy searches through Betsie.
Hunch access restored.
assoc. domains switched to producing a WML output.
Thesaurus button works, but found to be a bit heavy, and a good candidate for filtering.
Improved php lookup service, concentrating on function lookup instead of a general search. Pages still rather "heavy" though, and will need custom filtering. 'Possible that there are specialist mirrors that do this already.
Glitch with RIPE IP lookup fixed.
Issues fixed:
Mysterious spontaneous massive existence failure of the Icelandic server during mobile testing at Strange O'Clock.
Created site History section; split that and the legal section off from the main documentation.
The site Hosts noticed their server had fried, and switched it back on again.
Issues fixed:
As one of the results of the latest test & analysis, adjusted "flickr tags" search so it returns results less like "flickr recent". assoc domains blank button behaviour corrected.
Split unstable features off into "The Lab" (now disconnected from public view). Added Wapedia button (in the Lab).
Added the Status Summary to the main documentation.
Demoted Hunch to a link and promoted that link as Stable. Changed Wikipedia provider to Wapedia and flagged as Stable. Shifted position of some links around, to make more visual sense.
Experimental CSS to reduce the gap between the heading and the search field. Further CSS online for coloured buttons. Merged CSS Style test page into MAP Lab. Increased the font size of the search field.
Added lots of empty query checks and string modifier helpfulness.
Merged code for view source and head request (because they have to use each other anyway), got head request working (offline).
Took head request out of the Lab.
Added test taptu handler, researched (using the Portal on my mobile) the search FORM form, and moved the taptu button out of the Lab.
Lots of documentary updates.
Issues fixed:
Merged HEAD request into view source; uploaded new scripts, removed old test scripts.
Minor alterations to documentation pages during Validation.
Merged in gethead, partly as alternate mode, partly for 404 fix.
Some variable names shortened, very long lines split, indenting sanitised.
Filled in proper Server, filesize, and other HEAD grab info.
Well-formed XHTML output, with no open paragraphs now.
Additional link, leading back to the Portal for [filtered] Browse mode. Corrected spelling of comments &c.
Clickable href &c links within the document, initially absolute URLs only.
Tidying of code into a proper pseudocode section, although there is still much to do.
Separate offline scripts now working that detect different filetypes, and act upon them, other string-splitting things ready for mix-mode syntax colouring.
Added Texpedia button to the MAP Lab. Slimmer php function lookup, faster taptu, and unaltered-but-forgot to test-the-damn-thing possible facebook groups debug.
Issues fixed:
Wrote a "code mode", which displays non-text files. Yet to actually add that in, which is awaiting the addition of a more complex decision tree, itself waiting to be tested with a string splitter.
Fixed bug in the new link highlighting code in view source, wherein stray ampersands used to produce odd results.
Also changed those links so they view the source of the new file, rather than simply directly linking to it.
Enhanced comment-handling subroutine so that stray ampersands there are also handled elegantly, as an extension of the other stray ampersand bugfix. Removed part of the CSS fudge for this.
Proper pseudocode for view source, easing more complex expansion.
Wrote a plain text mode, or rather, made a test mode explicit.
Rewrote junk stripper routine: Knocking out windoze linebreaks, and odd characters, excessive whitespace, and general messiness, invoked when appropriate.
Combined the mode decider (already containing the junk stripper and text mode), and the code mode (still to be further developed as a separate script for the time being), into view source.
Modified to only offer to view with browse filter if file mode is HTML.
Now the following modes are available, and will select automatically:
-And these modes are faked, but will still be called:
Developed the string splitter to full readiness, and added that in.
This is easier said than done: I had to rewrite the rewritten code, but it's
working now, and token CSS & javascript modes are in. The format of the
code mode isn't out of tests yet, but the code mode is onstream
nonetheless.
The javascript and CSS modes are now ready to be developed into more than just text, which they are already, in a way. For the time being they just look pretty and highlight dashes.
No updates for a while, due to partial harddisk failure. Nothing vital was hit, but all resources were suddenly diverted to frantic mirroring, drive swapping, integrity tests, etc.
Getting ready to put Mavis in place of Betsie, ran into difficulties with the sexy new HTML parser not working quite as it should, so reverted to tag soup mode, which still played up with real world data, having performed flawlessly in tests. So Mavis is now on the back burner, along with all the filtering benefits it would have brought, and it may be simpler to try MODding Betsie.
The unhelpful labelling of wikipedia as textpedia, and wapedia as wikipedia has now been dropped, because that was just getting too confusing. Both options now next to each other on the main page.
Elaborate attempts to get monospaced text to behave in code mode have been defeated by senseless mobile browser design, so now it will make a best attempt, but some things might not line up. This is still better than everything definitely not lining up, as would have been the case using the complex methods which work on desktop browsers.
The problem of some coloured text coming out white has been solved; it was a bug in some mobile browsers intended to prevent low-contrast colour combinations being used. Solution: Use a slightly lighter colour.
The advert and feedback options are also being quickly redesigned, with an eye on simplicity. The feedback system is unlikely to come under complex spam attack while it is in it's current form, therefore automatic DoS detection has been dropped from its brief. In the event that it does get attacked, that can be noticed and corrected manually, with the price of a one-day delay.
The MAP Lab is being wound down in scale, and may be closed in future. Irritating behaviour from Mavis may delay this. CSS Experiments with buttons are for nought: Either browsers display buttons consistently or they don't, and those that don't can't be worked-around. Facebook's method of removing the blue border from active buttons turned out to be a fudge: They just made the background of their buttons the same colour of blue as the borders. The dark colour also hides their 3D-ness. This method doesn't work very well for yellow buttons, which look fine anyway on compliant browsers, and passable on the others, and so that seals the fate of the "Style Lab" button test area.
Issues fixed:
Switched the new Category List & Popular caching routines on.
Renamed combined.php to myhunch.php, the new stable name, which it will remain inhabiting for the foreseeable future. Added global references to callable scriptnames, to allow for easy renaming in future.
Took the MAP Lab out of public view. This is because there is hardly anything useful to users left in it, and this was starting to make it look "messy". It still exists and functions, and will still be referred to, though.
Streamlined internal standards a bit more.
Modified display of results.
Unfortunately, percentage results do not show due to a bug, but that bug causes no actual harm.
Added "popular" behaviour as a highlight result, but either another bug had got 'em, or I haven't run across any to highlight yet.
Implemented TransWiki; hunch.com Wikipedia links are turned their wapedia equivalents, and for future-proofing, this works across the language barrier, too. (Or at least the country code.)
Changed behaviour of clarify mode when there is no query, to a welcoming home page, with as much of the style ported from the desktop hunch as possible.
More than doubled the number of operating modes, bringing in test routines to the main script.
Took mobile hunch out of the MAP Lab and into Alpha!
Right, that's that done: Now off for a short break and some sparkling espresso...
Noticed that although I'm ready to upload when the local connection points are closing, hunch's offices on NYC time have four hours more to go!
So it's off down to bath, to continue.
Some of this may count as bugfixing; I'll check on that shortly... Nope; no bugs there! Hardly any bugs to speak of at all, infact. -Ever onwards...
A few updates before then... 'Will document later...
Upgraded the offline server to php5 and proper cookie handling, and got Python working on another machine.
Added the Quote All tickbox option.
Upgraded the advert code, making it easier to add/remove/edit adverts
in future.Added whitelist & blacklist, for preventing abuse of the feedback option, and separated it from the rest of the code, in much the same way as the advert handler.
Reinstated the Thesaurus button. Now the MAP Lab has nothing to do until we find out where the TV and Bus times are stored online. Phone numbers would help too. Some de-cluttering may lie ahead.
Tested adding the feedback option. (Still under test as I this was written, but worked well enough to debug online!)
Visual rearrangement of tickboxes so they don't "lose" their text.
Made Quote All unticked by default.
mobile hunch now strip's slashes from queries, ie not of \"this nonsense\".
Issues fixed:
Removed php4 compatible JSON code from mobile hunch and (the unreleased) version 2 of the search filter demo.
Boolean controlling page and links to begging bowl feature, enabling it to be turned off in future, if required.
Handling of API Keys shifted into the call handler from calling routines, to make automatic key handling easier to switch to later on. &/? URL-joining bits included in that, too.
Some test server strings were titled $fakeX. Changed to $fauxX to avoid accidentally insulting Catherine Fake in future.
Replace included occurrences of called filenames with a self-detecting variable to prevent scripts possibly switching over at some links in future.
Gained formal approval from CEOP to use their button
Finally got around to implementing the CEOP button on the portal, delayed "slightly" by the login wrapper for mobile crunch driving me round the twist and being regurgitated in various other projects to keep it's development "alive", having to do a wide variety of other work, and of course juggling funds as usual.
Sudden revival upon finding a way to properly communicate with Google, by accident. define Button added. Not tested properly yet though.
Here live legal notices explaining who runs which service, and clarifying that we're not trying to pass ourselves off as them, violate their TOS, steal their tea towels, etc. It is worth noting that we do provide some of the services ourselves as well, such as the source code viewing feature, and one of the filtering options.
The general idea is to make life easier for mobile users. Traffic is passed onto third-party service providers as a side-effect of this.
This service has recently been rolled back to Alpha test status. Aside from the practicality of not jumping straight into beta before the bulk of the service features have been tested, released, or even written, TOS's may inadvertently be technically broken if a commercial service is running, which it can't fully be if it's still in alpha. Furthermore, it is not practical to ask for clarification of minor points of other parties' policies, in relation to demonstrations which aren't available yet.
A list of services, who provides them, how information is passed between you and them, and the current legal status of this exchange.
The plain links are simple hyperlinks to email providers' mobile login/homepage screens, and work through directly through your browser. They are, and are provided by, Hotmail, Google gmail, and Yahoo mail.
The last plain link is special. It look's like a button, but is infact still a link. It has to resemble a button for legal reasons. This is the CEOP button, which links to advice on child protection issues. Because CEOP's homepage's link to their mobile pages does not currently work on mobiles (!), this links directly to their mobile advice. It has gone through CEOP's formal approval/registration process.
The informative text only provide towards you, and do not use or store informative from you in any way. They are all provided by us, directly. This text consists of links to various portions of this documentation page, and what time it is (or was), and what the day and date currently is (or was). Links to various adverts are also served.
History & Legal Contents
Main Contents
Back to the portal, or the main documentation