Now, I’m not particularly surprised – this was always likely to be the endgame. What I am surprised by is the fact that there doesn’t seem to be a coherent post-flex story coming out of Adobe. There are mentions of MXML/HTML5 ‘conversion’, but no information other than that. What you’re supposed to do in these scenarios is to present your next-generation tools at the same time as deprecating your existing toolset. I think not to do so is a bad strategy for Adobe.

Stepping back for one moment – why do we use Flex at all? It clearly has disadvantages; but despite the anti-flash sentiments from the peanut gallery (who may well have never used a flex application in their life), it actually represented a good choice for a startup business building intranet targetted applications for customers who were tied to IE6. Adam Flater sums these advantages (and disadvantages) up very well in his blog post:

The main benefits of Flex in it’s prime:

• A standard user interface component set
• Remoting (the ability to interface with web services via transferring typed objects)
• A better skinning and styling workflow (than HTML/CSS at the time)
• Efficient vector graphics for data visualization (charts, graphs, etc)

We’ve found it a relatively pleasant way of constructing applications – certainly it enables us to be very agile in changing how the UI works, even with a small development team. I often wonder if the ‘you should use jQuery / HTML5 / the new Hotness This Week’ crowd have ever actually written a web UI at “application” scale, rather than believing that a single-developer Web2.0 ‘todo list’ is representative of the challenges building full-scale applications present.

Back when we started 4+ years ago, we looked at the available alternatives. I hate (and still hate) Javascript. I liked GWT, I just thought Flex was a better choice for us. And just because Adobe aren’t going to make any more of it, doesn’t mean we suddenly have to find an alternative right now any more than the fact Ford don’t make my model of car any more. I can still drive to the shops. I can still put fuel in it. I can still buy spare parts. All that’s happened is I now know how many miles are on the clock.

Indeed, I can understand why Adobe are probably doing this – there’s no money to be made here. Unfortunately the timing and poor message has probably just blown any opportunity of making money in the future, because they’re moving from a “Flex tools > HTML tools”  world and into an “Adobe HTML tools > everyone else’s HTML tools”. But the signs there don’t look too great. CS5 sucks; the UI is an unusable mess. Dreamweaver is awful. Flex builder is ‘adequate’ at best.

I think there will be a lot of knee-jerk migrations that happen very quickly. Myself, I think we have about 18 months to decide on what the next generation platform is to build on. My early guess though : Google’s Dart. If you watch this talk about Google’s closure tools, and then think about Dart it makes sense as a strategy. I look at it and intrinsically feel like the problems they have are the same as the problems that I have.

So, if I were Adobe, I’d have announced a complementary product that did AS3->JS compilation (such as Jangaroo), a rich HTML5 widget set, and libraries for AMF remoting. Dart is pretty immature right now – AS3 basically has all the features that people would want. But I have a feeling that by the time Adobe finally work out what they’re doing, the advantages that Flex had over HTML5 will be solved by other, free tools, and their customers that were on flex will already have migrated to something else.

What was more irritating was that much of the music I’ve already got is sitting on CD in lofts in buildings where I’m currently not. Much of it is also on YouTube of course, but it’s not as pleasant to access.

What to do? Well, it’s well known that Audio Hijack Pro can rip the audio from Spotify. But who wants one giant 10-hour MP3 to deal with?

Sitting around with some time on my hands due to a broken hip, I cobbled together a script when I found that groovy was capable of invoking AppleScript scripts. Taking 240mg of codeine / day was not a time to start learning how AppleScript does file handing; this is all pretty hacky stuff and could be improved on, I’m sure. Think of it as proof of concept.

Steps:

1) Get groovy installed

2) Turn the debug logging on in growl

3) Start hijacking in Audio Hijack Pro

4) run the following script (e.g in groovyConsole)

5) Play music

The script will tag the music you play, and split into individual files. It probably won’t be happy with things like “s in track names, but fixing that is left as an exercise for the reader. Please bear in mind this was written literally on drugs.

import javax.script.*;

class Processor {
    def f = "/Users/magnayn/Library/Logs/Growl.log";
    def process;
    def engine;
    def run() {
        def mgr = new ScriptEngineManager();
        engine = mgr.getEngineByName("AppleScript")
        startRip();
        process = "tail -0 -f $f".execute();
        def lastLine = "";
        def inSpotifyGrowl = false;
        def spotifyLine;
        process.in.eachLine { line ->
            def idx = line.indexOf(" Spotify: ");
            if( idx > -1 ) {
                spotifyLine = line.substring(idx + 10);
                inSpotifyGrowl = true;
            } else {
                if( inSpotifyGrowl ) {
                   def line2 = "";
                   if( !lastLine.substring(0,10).equals( line.substring(0,10)) ) {
                        line2 = line;
                   }
                   trackChange(spotifyLine, line2);
                }
                inSpotifyGrowl = false;
            }
            lastLine = line;
        }
    }

    def trackChange(line1, line2) {
        try {
            def album = line2.substring(0, line2.lastIndexOf(")") );
            def artist = line1.substring(line1.indexOf("(")+1);
            def title = line1.substring(0, line1.indexOf("(") ).trim();
            rename(title,artist,album);
            println "Title: " + title;
            println "Artist: " + artist;
            println "Album: " + album;
        }
        catch(Exception ex) {
           ex.printStackTrace();
        }
    }

    def startRip() {
def scr = """
tell application "Audio Hijack Pro"
  set theSession to first session whose hijacked is true
  start recording theSession
end tell
""";

engine.eval(scr);

    }

    def rename(title,artist,album)
    {
def scr = """
tell application "Audio Hijack Pro"
  set theSession to first session whose hijacked is true
  split recording theSession
  tell theSession
    set title tag to "$title"
    set artist tag to "$artist"
    set album tag to "$album"
  end tell
end tell
""";

engine.eval(scr);    

    }
}

def p = new Processor();
p.run();

So, when I bought a new Macbook Pro for work, I didn’t mind so much that the new video connector was mini-displayport, thus obsoleting my drawer-full of macbook->{various} display converters, because the 3008WFP comes with a kosher displayport connector, that was of no earthly use when I bought it, but means I don’t have to buy Apple’s *very* expensive mini displayport -> Dual Link DVI converter.

So, I bought from ebay a 3-in-1 mini-displayport to HDMI / (single-link) DVI and Displayport converter. The DVI is only good for 1920×1200 resolution – you need either dual-link DVI or displayport for the full 2560×1600. So I plug the connector into it, use the (very robust looking) dell displayport lead and….. 640×480, maximum. Gah!

Cue much faffing with SwitchResX, learning about display EDID settings and… no further forward. There’s forum postings elsewhere of similar issues, sometimes with corrupted EDIDs, but this wasn’t the case here.

Once, and once only, did the mac offer me 800×600 – I thought it might have been a fluke or a magic setting after playing with DisplayResX, but I could get no further. Sigh. Perhaps the Dell implementation was too early to be properly compatible.

On a last resort, I tried a new, single mini-displayport to displayport adapter – (not very expensive from Lambdatek). Aaaand – boom – full resolution offered. The 3-in-1 is obviously a bit too shonky to work at the resolutions required.

So – if you’re googling with a similar problem and have gotten to here – it’s worth trying a new cable !

First of all, erstwhile Google CEO succumbs to foot-in-mouth syndrome, and says (in reference to your usage of the internet and Google searches in particular) that “If you have something that you don’t want anyone to know, maybe you shouldn’t be doing it in the first place.”

Now, what he was probably trying to point out was the fact that ISPs and providers these days do keep transaction logs that can be forced to be handed over to “law enforcement” (I use the term in the loosest possible sense). But it doesn’t come across that way. This is the Schmidt, after all, that Blacklisted CNET for publishing things that were discovered from…. Google searches. Hypocritical in the very least.

But the fallout continues; the Google honeymoon is clearly over. Firstly, Bruce Schneier pwns the argument being laid out (legal, proper, unembarrassing today != legal, proper, unembarrassing tomorrow); what you’re prepared to put on display as an 18-year-old party animal might not be the same as when you want to be a middle-aged “pillar of society” (just look at the problems Messrs Cameron & Johnson have had over Bullingdon Club photos – and that wasn’t even from the Internet Age…).

Next, I wake to a feeling that I’ve been transported to a parallel universe, that an exec in Mozilla effectively comes out to say Evil(BORG) < Evil(GORG), so Here’s how to switch your default search to Bing. Maybe he’s a privacy nut, but Mozilla make a majority of their revenue through google searches – so he’s at the very least a brave nut.

The perception of Google has already tarnished, and not because of the latest privacy debacle. Google is currently blundering about like the Incredible Hulk, “having fun” they say, seemingly oblivious to the collateral damage it causes. Consider: Microsoft some years ago was sued by the US Government for bundling Internet Explorer with Windows, for free, thus unfairly restricting the market for competing web browsers. MS argued that web browsing was such a critical component of an “operating system” that the two were inextricably linked. Either way, this probably killed Netscape, and certainly no sane company was likely to pursue a strategy of browser product sales. Fast forward to this year. Google announce “free turn by turn maps for android”, because “maps are such a critical component of the mobile experience”. Pattern? Netscape == TomTom? Google effectively cross-subsidise their crushing of random markets with money from their monopoly over internet advertising. Sounds pretty, uh, anti-competitive to me. How long before the hulk blunders its way through your business district? Maybe you can scrape a paycheck out living in Google Opt-Out village…

Anyway; I digress. For many, the real privacy issue isn’t about “the government”. It’s largely about “non-friend” people who know you. For the non-famous amongst us, that’s casual acquaintances, colleagues (both subordinates and superiors..), friends-of-friends, people at work, down the pub, etc. For the famous of course, that’s a set that expands to fill most of the world. And it’s about perception and reputation management.

I think the reason Facebook became so successful is that it, wittingly or not, understood this. Without such a site, what are my choices for posting, say, photographs of an office party? Well, there’s flickr, a blog, a webpage. Hmm.. not *too* keen on that picture of photocopier abuse going “general release”. So a closed flickr page perhaps – mail round invites? Hm – not very appealing – I don’t have addresses to hand – too much effort. Send by email… ugh… spam people who may or may not care?

No – the selective privacy part was fantastic. And I say “was”, because the Latest set of changes make this significantly worse, and show a clear direction that they intend to persue – one of trying to coerce their users into “everyone” visibility. For them, that means more free content, more ads, more profit. For me? Well… I can’t help imagining the scene in a few years time. “Thankyou for your application for CTO of GlobalMegaTech, Inc. Can I start the interview by asking if you think being a ‘fan of Iced Gems’ is appropriate for a senior position in a major multinational?”. Or “Hi Dr Jones. Before we talk about my lung transplant, and you give me the smoking lecture, can I ask about your fandom of Benson & Hedges?”.

Yeah, even if they don’t ask it, it’s out there. Luckily I’m not famous, or work in an industry or jurisdiction that is known to be particularly humourless. But perception is important (ask Max Mosley), and the change to force “Everyone” to seeing pages is dumb. Part of the fun was that you could persue small acts of whimsy amongst selected friends in the “global village” without the perception fallout to the wider world. Is that, after all, the very essence of being “Social” ?

I’ve already had one friend divorce facebook; I suspect because of the problems of How Your Creepy Ex-Co-Workers will Kill Facebook. I don’t want to go that far – it does still provide a point of contact that isn’t yet replicated well by other services; but I’ve removed all information that you can’t hide from “the public”, and I can’t see me using it much more. Ironically it seems that Facebook wants now to move more towards the things that it replaced – instead of controlled privacy, a “Geocities” of content exposed to all. A social network for the Proletariat, for those who photocopy their arse at the xmas party and don’t care who knows it.

What’s a shame, from a technological standpoint, is how centralised all of this is. Networks rise, networks fall – for the reasons linked above – but in this instance it’s because a centralised party decides to change the rules of the game (as is their entitlement). What’s needed is a decentralised system. All the technology exists – PK certificates and countersigning neatly give you Friend-of-Friend access control – what’s missing is the drive to ‘make it happen’ – to wrap up the technology into something user-friendly, user-deployable and user-usable in the way that Facebook is, and OpenID, PGP and the rest of the geek-tech are not. A way to put up facebook-like social sites accessible globally, 24/7, but visible at a granularity of your choosing.

So far there has been no catalyst. Maybe a Facebook Terms-of-Service change will cause it to happen (geeks after all tend to care disproportionately more about their privacy). My guess is probably not. But there may be a catalyst coming.

Switch to P2P for a moment. Their developmental history is one, predominantly, of laziness. Napster was king for a while – easy to use, not terribly complicated, but it works, so why bother expending energy on something more sophisticated? However, it was an easily attackable system (single, centralised server!), so it eventually got crushed. BitTorrent then sprang up – but still identifiable through clients connected to a tracker. Peter Mandleson wants to “crack down on this sort of thing”. Clearly he’s never seen The Sorcerer’s Apprentice. We already know what the next-generation will look like – Microsoft wrote about it in 2002 – the Darknet paper; decentralised, closed, interconnected networks. The technology is already out there (e.g Freenet), but they’re not what you might call “consumer grade”. But nobody can be bothered much to fix the UI because… well, there’s BitTorrent, right? Necessity is the mother of invention.

The connectivity (privacy, authentication, location, discovery) issues for such a P2P darknet look a lot like the connectivity issues for a decentralised social network – How do I find people, how do I maintain and control privacy, how do I share things.

By trying to strike down the current “good enough” BitTorrent, Mr Mandleson may just provide the catalyst to make it happen.

Random reads are very, very fast. It really doesn’t care about fragmentation; apps start up pretty instantly. For developers with a large working set (with all the code intellisense guff in Eclipse et al) it’s a significant benefit. You loose a lot of the ‘oh my god what IS it DOING I’m going to THROW this machine out the WINDOW’ moments. I STILL wish Eclipse had a clue about FSEvents, but you can’t have everything. It’s surprising how slow going back to a spinning platter feels when launching apps.

Now the bad news. Proviso: The way I use a disk might not be typical for an end user, but is probably pretty typical for a Java developer. It’s plugged into my 2008 Mac Pro, on which I do a lot of building of a reasonably big application, which constitutes a maven build (or an eclipse background compiling). So a lot of .class files and .jar files will be created and deleted every day. I do however leave a fair proportion (~20%) of the disk unused to try and give some headroom for defragmentation.

Now, this may be the worst case scenario for an SSD, but I’ve noted a couple of blog posts of late about slowdowns on heavily-used X25-Ms (e.g SSD Reconditioning requiring HDDErase (or equivalent) in order to coax them back into their original performance. For me, I don’t know if it’s the combination of heavy dev work, or an installation of Snow Leopard over an old install causing excess fragmentation, but my machine had slowed to an utter utter crawl, locking up in several places under load.

Some observations:
* HFS+ totally sucks, and needs to Die. There’s a build plugin that I use which creates many temporary files in /tmp, and when it cycles through deleting them on completion, I/O effectively locks solid. This is really poor. Something’s rotten in the state of OS X I/O; VMWare also seems to be able to bring the system to its knees.
* the X25-M is hardly painless for Mac users, you need a PC to hand to upgrade the firmware (if you happen to have an NVidia graphics card?!?) or to run HDDErase to refresh it (oh, using v3.3, not 3.1 or 4.0, and on a PC with a specific BIOS… FFS!). I can cope with a “deep clean” cycle once in a while for the benefits, but it’s a PITA to have to get the screwdrivers out in order to do it.
* Carbon Copy Cloner is your friend.
* TRIM is supposed to help (if the OS actually supported it). But Intel says no TRIM for G1 owners (Well, actually, Intel officially says _nothing_ – they must be learning from Apple) – fair enough, but it’d be Kudos to them to support the early adopters. If the firmware just won’t fit in the available flash space (!) or whatever, then just say so.
* There’s a lack of a decent, mac, free, HDD performance tool. XBench isn’t accurate. Someone should bring IOMeter into the one-click generation.

So – a qualified thumbs-up, a bit more so for business expenditure, a bit less so for personal expenditure.

So, I have a messy desktop, and a messy desktop. I develop software much in the way a chimp with Attention Deficit Disorder might – closing windows is for losers, and if the compile takes more than a minute, I get bored and start running up web pages and generally tinkering. Which gives the hardware a hard time – my working set is probably *huge*. Which in turn is a problem as I end up shouting at the slowness of the environment.

Hopefully, the answer may be the X25-M SSD. I had ordered one, that failed to get fulfilled by some dodge-o vendor; and I then got worried by the ‘it slows down’ internet pages. In the end, I just got too impatient – the only other game in town was a 300Gb Raptor – and that is sufficiently expensive to mean I couldn’t really justify that *and* an SSD in the near future. So I re-ordered, and helpfully Anandtech have a newer review that says ‘yeah, they slow down, but still trounce magnetic media even then’. So, here we go.

Sidenote : If I were an expensive database vendor, I’d be worried. Consider – part of the database proposition is that given

* Dataset size >> Memory
* Memory speed >> disk speed
* Disk Latency >>> Memory Latency

therefore you spend $$$ to have some clever code try and mitigate these factors. But consider – our current app generates about 1Gb of data per year. It barely needs a database, let alone an expensive one. And if the disk were quick enough, merely serializing object graphs to disk becomes a viable option…

Unfortunately, every time you save the project, the order of references in this (XML) file are randomised. This plays havoc with SCM tools, as you basically get massive conflicts every time, as it looks like the file has been completely changed.

This is annoying.

Fortunately, because git is wonderful, you can apply ‘clean/smudge’ drivers to pre or postprocess any files in your system. Time to crack out a bit of script, to re-order the files into something more sane..

Assuming you have all the groovy gubbins installed; Put this in ~/bin/fixFlexLib.sh :


#!/usr/bin/env groovy
def lib = new XmlParser().parse(System.in);
lib.includeClasses.each {
ic -> ic.children().sort { a,b -> a.@path <=> b.@path }
}
lib.includeResources.each {
ic -> ic.children().sort { a,b -> a.@destPath <=> b.@destPath }
}
def writer = new StringWriter()
new XmlNodePrinter(new PrintWriter(writer)).print(lib)
println(writer.toString());


Then, append to your ~/.gitconfig

[filter "flexLibFilter"]
clean = ~/bin/fixFlexLib.sh


Finally, in each errant flex project, add to .gitattributes in the root (or wherever), not forgetting to add it to the index:

.flexLibProperties filter=flexLibFilter


Job done.

I remember when it arrived as our saviour from the horrors of Visual Age for Java. I’ve used it ever since. But now I’m reaching the point where I wonder if I should just call it a day, and ship off to IntelliJ. I’d like not to, as there’s a lot of muscle-memory tied up in keybindings.

But I may have to. I’m spending too long shouting at my computer wondering WTF Eclipse is _doing_.

Now, I do not have a slow computer. It’s a monstermac, 8-cores, plenty of RAM, 10K RPM Raptors.

So, firstly, examining a static eclipse, it’s using *795* Mb of RAM. Doing *what*? Garbage-collection frees memory for all of 10 minutes. I could run JMap / JConsole, but – oh yes – that’s a 1.6 JVM thing, and we’re stuck in eclipse land on the Mac in 1.5 because – well who’d have predicted it – 1.6 is 64-bit on the Mac and there’s no 64-bit Carbon. Not like Apple have warned about this for – ooh – _years_ or anything. Oh yes, some half-baked SWT implementation is in 3.5M3, but *really*.

It baffles me. Is nobody actually using this thing in production? Is it even being tested? My install burps up errors like they’re going out of fashion. A selection :

Error: Cannot resolve Platform symbolic name mac os x:i386

Error: /Applications/eclipse3.4/configuration/org.eclipse.equinox.source/source.info (No such file or directory)

Error: Cannot resolve Platform symbolic name mac os x:i386

Are these things important? Should I be worried?

More bad things:

The ‘X’ errors randomly stay on my projects, even though they’ve built correctly. Or, at other random times, it arbitarialy announces that java.langObject can’t be found and the build path is inconsistent (but it isn’t), which cues the random refresh->clean->rebuild->refresh->AAARGH cycle which sometimes works. The Workspace is *permanently* out of date. I’m sorry. I can’t live in a hermetically sealed Eclipse environment. I need to edit files in BBEdit occasionally. But don’t click ‘refresh automatically’ as you’ll never see your eclipse again because – and this baffles me – *Windows* is the only platform with OS-level change-eventing support. No FSEvents on the mac. No inotify on Linux. You’re SOL, just grindy grindy disk chomp for you.

To quote The Shatner : “None of this shit works for me”.

Unfortunately, everything else sucks. Badly. I could rant all day about how bad the user experience is, but to sum it up, let’s look at networking.

Now. On my iPhone, you want to use WiFi – you tap a couple of times, maybe have to type a WPA Key in, and that’s it.

On Wince. Well… on this HPAQ, there’s the ‘Connections’  tab in ‘Settings’. So far so good. But in it there are several seemingly possible options : Connections, Wi-Fi, Wireless Manager and HP iPAQ Wireless.

Want to guess which one of those is the right one? Bzzt. Wrong. You can connect with any of them. IF you have the right Fu. May I suggest that you do not click on “Wi-Fi” and wander anywhere near ‘Advanced’ as you get a dropdown and a terrifying list.

My network card connects to: (Work or The Internet). Huh? Both, as it happens. I actually know why it’s asking, and that’s also a big problem.

And then – and I wish I were making this up – there’s a list that invites you to ‘Tap an adapter to modify settings’. Here’s the things you could tap:

• NE2000 Compatible Ethernet Driver
• AsyncMac NDISWAN Miniport
• Remote-NDIS Host
• Virtual Ethernet Intermediate Miniport
• Microsoft null Ethernet Miniport Driver
• L2TP NDISWAN Miniport
• PPTP NDISWAN Miniport
• Marvell SDIO8686 Wireless Card
• Bluetooth PAN User Driver
• Remote-NDIS Host

Now, I’m old enough and sad enough to have an inkling what some of that means – God help anyone that might possibly view it as a tool to – you know – get stuff done. Arrrr! There be settings to be broken here. 

My Mum wouldn’t know what that meant. And nor should she. If Microsoft really, really think Windows Mobile has a chance against vendors like Apple without starting again, from scratch, they’re either shortsighted, stupid or both.

To wit:

Recomac Surfacing Limited

KPMG’s joint administrators … offer for sale the business and assets of Recomac Surfacing Limited. Principal features of the business include:

• Annual Turnover of circa £6m, split 80/20 between the surfacing and sports sectors of the business
• Recognised as a market leader in the field of highway maintenance and sports surfacing
• The company has an extensive customer base in both the public and private sector
• Plant and machinery includes a fleet of 20 bitumen sprayers
• The company operates from a 12,381 sq ft freehold trading premises in North West Manchester

Sounds great. What could possibly have gone wrong? Roads always need fixing. Olympics round the corner, big sport public expenditure? Heeeeeeeere comes the kicker :

• In addition, the Company has for sale a development of 30 two bedroom apartments within a 15 minute drive from Manchester City Centre. The apartments are fully complete.

I can see how that was in total alignment with your main business.

FAIL.