�t�p�n Roh - Alive But Sleepy

Confessions of a Developer

Sun, 08 Jan 2012 22:47:00 CET

It was fall of 2000 and I was freshly kicked off the University and broke. Being broke I needed to get a job quickly so I turned at this software development house my friend started a few years earlier. (I was so broke that I had to ask for advance paycheck before even working a single day).

A Simple Task

I was immediately assigned a simple task - create a Multimedia Digital Archive. In Java. As a sole architect and developer. Bear in mind that I had one semester of Java, wrote in it one application which was in fact a port of my older school project, a web server written in C. I remember thinking: "What a weird language that Java, there are no global variables!" My only qualification was that I did a rewrite of MDA's Perl predecessor as a summer job a year before.

To be honest I did not work on something the size of future MDA - the Perl thing I worked on was just a tool pulling text and images from various sources, normalizing them and storing them on a disk. Two PHP applications then allowed to browse, search and modify them. The requirement for new MDA was simple: replace the Perl tool with a real server, but keep those two applications compatible,

Architect is Born

As a freshly appointed architect I knew I had to proceed in a right way. So I fired up Dia and designed the whole system as a huge class diagram. That felt great and I was ready to get dirty. Java IDEs were huge, slow, often crashing and ugly looking back then so I used my trusted too-sexy-for-syntax-highlighting Joe. In a no time I had the server written and running. It had everything - socket server with telnet-like interface to be used by PHP front-end, whole-in-memory highly optimized set of B-trees as an index and Perl as a scripting language.

XML Rules!

As time quickly passed, we realized that having PHP front-end is not satisfying anymore and devised a plan to implement web UI in Java. Being totally ignorant of what goes on in the world of enterprise Java, I went past all web frameworks and chose the basis of them all - Java servlets. Combined with a new favorite of ours - XML. Data were exported as XML by the persistence and made into HTML by XSLT. It was very elegant, took a long time to master, was hard to extend (new report page? I don't think so) and slow as hell. But worked.

However, the road did not end there, the biggest thing of them all was yet to come - we needed a rich client. By that time I made a peace with Java IDE named NetBeans and used that for GUI design and coding. That solved the problem of designing UI without seeing it, but I was facing bigger problem: how should client and server communicate?

The Abomination

Telnet-like interface did not seem modern, especially when both parties were coded in Java so I turned to our trusted technology - XML and created the single-most--words-cannot-describe-it-properly--remote-protocol-in-the-world: remote DOM. I'll wait for you to process that… I can wait… OK, you're back. DOM is better known at the browser side of the web where it seems to serve its purpose. However in the server world, it is a terrible nightmare to work with. It is not Java-friendly, uses strange patterns and consumes a lot of memory. And I made it working across the network. And I don't mean that it supported a subset of it. No, no, no. It was fully working, remotely-operated DOM with all those Nodes and Elements and Attributes and who-knows-what-else.

Epilogue

Fortunately MDA got replaced later and company is now defunct so noone can experience for themselves what 22 year-old developer was capable of and what still haunts him in the sleep.

GomokuJS - Play Gomoku at home, at work and on the road

Sat, 20 Nov 2010 17:02:00 CET

GomokuJS is a little web-based game of Gomoku. One can play against computer (I admit that not very bright one), against another player on the same computer and even over network with a friend.

PLAY

Supported are recent versions of Mozilla Firefox, Google Chrome, Opera, Internet Explorer, Safari, Konqueror (with few visual glitches), iPhone/iPad, mobile devices with Opera Mobile (tested on S60-based Nokia E71) and Android-based devices (unfortunately without network play).

A Little Bit Of Music

Sun, 30 Aug 2009 16:08:00 CET

Queen medley played by Harmoniemusik Rohrdorf recorded on July 4th at Filisur.

Search and Comments (and Twitter)

Tue, 16 Jun 2009 22:24:00 CET

Web 2.0, here I (slowly) come. Comments, trackback (powered by HaloScan.com) and site search (powered by Google) are now just a click away.

(Day later.) Added Twitter box on top (where, besides other stuff, server status of srnet.cz will be published using hashtag #srnetcz).

Pr�ce v Polarionu

Sun, 14 Jun 2009 23:06:00 CET

Aktualizov�no.

Odd�len� v�zkumu a v�voje Polarion Software s�dl�c� v Praze sh�n� v�voj��e v Jav� a pracovn�ky technick� podpory. Alespo� z�kladn� orientace v oboru a dobr� znalost angli�tiny nutn�. V�ce informac� poskytnu na po��d�n� mailem.

shttpd/0.3.1 - give credit where credit is due

Sun, 14 Jun 2009 20:46:00 CET

The reason of this blogpost is to give credit where credit is due: Rimmi Devgan and Avanti Limaje, two MS students with the Computer Science Department of the State University of New York at Stony Brook found a serious security vulnerability in my little web server software called shttpd. I wanted to thank them in public (as I call this dark and lonely corner of the web).

The release made it to the Root.cz, online Czech magazine about Linux. Have no idea why. Maybe because I made the first release in 9 years, probably the record in history of software. Or maybe not.

Unreliable System.nanoTime()

Thu, 17 Jul 2008 22:08:00 CET

Java's System.nanoTime() is better than System.currentTimeMillis(), right? Wrong!

There is a subtle price paid for better accuracy of System.nanoTime() - on various platforms its value is taken from various sources. These sources may sometimes fail. And if they fail, garbage gets returned (most probably zero, but who knows for sure). This may seem like a very rare situation, but combination of hypervisor Xen and Windows produced garbage too much for our liking. Maybe it was caused by Xen bug #1146.

So if you measure something with System.nanoTime(), check whether the measurement is bigger than zero and (as we also did) whether it does not differ too much (like 1 second) from the measurement done using the good old System.currentTimeMillis().

Navigace Garmin a Linux

Wed, 09 Jul 2008 16:30:00 CET

P�r rad, jak z Linuxu pracovat s navigacemi Garmin, konkr�tn� s typem Garmin eTrex Legend HCx.

P��stup na microSD kartu

P�ipojte navigaci dodan�m kabelem do USB portu a zvolte Hlavn� menu / Nastaven� / Propojen� / USB diskov� jednotka. Micro SD karta bude vz�p�t� v Linuxu rozpozn�na jako USB Mass Storage tedy stejn� jako nap�. USB flash disk. Je mo�n� kop�rovat data z karty i na ni.

Mapy a dodate�n� mapov� podklady jsou v adres��i garmin, pro�l� trasy ukl�dan� na kartu jsou ve form�tu GPX ulo�eny p��mo v ko�enov�m adres��i (pojmenovan� stejn� jako v navigaci).

Navigaci uvedete do p�vodn�ho stavu jedn�m stisknut�m tla��tka pro zapnut�/vypnut� p��stroje.

P��stup do vnit�n� pam�ti

Navigace m� je�t� vnit�n� pam�, do kter� se ukl�daj� pro�l� trasy, zna�ky, cesty atd. Tato pam� se d� zp��stupnit utilitou GPSBabel. Je mo�n� vyu��t dva zp�soby spojen�: p��m� USB a emulaci s�riov� linky. Standardn� po p��pojen� navigace USB kabelem se v m�m syst�mu nat�hne modul garmin_gps, kter� zprost�edkov�v� emulaci, tak�e jsem prvn� mo�nost ani nezkou�el.

K z�lohov�n� cest (routes), pro�l�ch tras (tracks) a zna�ek (waypoints) poslou�� takov�to vol�n�:

gpsbabel -r -t -w -i garmin -f /dev/ttyUSB0 -o gpx -F internal.gpx

/dev/ttyUSB0 je jm�no za��zen� emulovan� s�riov� linky p�i�azen� syst�mem p�i nahr�n� modulu.

V�sledn� GPX soubor je mo�no nahr�t do MapSource (viz n��e) a v�e si prohl�dnout na map�.

�daje je mo�n� i zp�tn� zapsat do navigace.

MapSource

MapSource je mapov� aplikace dod�van� k navigac�m Garmin, kter� umo��uje nahr�v�n� a prohl��en� map a pr�ci s trasami, cestami i zna�kami.

Instalace p�es Wine je trivi�ln� a aplikace funguje naprosto bez probl�m�, samoz�ejm� mimo p��m�ho propojen� s navigac�. Jedin� v�c, co nefungovala byl automatick� update na nov�j�� verzi, ale ru�n� spu�t�n� update z webu pro�lo bez probl�m�.

P�es Wine lze nainstalovat i jednotliv� mapy, nap�. dod�vanou TOPO Czech 1.20 �i Atlas Czech 7, kter� je k dispozici v�em vlastn�k�m mapy TOPO Czech. Mapy je t�eba odemknout 25ti m�stn�m k�dem, po�adovan� GPS Unit ID lze naj�t v navigaci v Hlavn� menu / Nastaven� / Syst�m / [MENU] / Verze softwaru, s�riov� ��slo GPS pak v prostoru bateri� �i na krabici.

POI

POI (Points Of Interest = body z�jmu) lze st�hnout z n�kolika m�st na internetu, nam�tkou POI plaza, GPS-Waypoints a poi.cz.

POI Loader sice lze p�es Wine nainstalovat, ale odm�t� ��dn� fungovat. Na�t�st� lze pou��t v��e zm�n�n� GPSBabel:

gpsbabel -i VSTUPN�_FORM�T -f SOUBOR_S_POI -o
garmin_gpi,category="KATEGORIE",bitmap=SOUBOR_S_IKONOU -F
V�STUPN�_SOUBOR.gpi

Navigace Legend HCx kategorie podle v�eho ignoruje (v�echny POI jsou ozna�eny jako u�ivatelsk� body z�jmu), ale jin� modely ji mohou podporovat.

V�stupn� .gpi soubor se nahraje na microSD kartu do adres��e garmin/poi.

Nahr�v�n� map

Kvalitn�m zdrojem ov��en�ch map je fatbozz.towerofglass.net. Pod Linuxem p��mo funguje popisovan� postup pro konverzi map z OpenStreetMap.

Pro pou�it� v navigaci je t�eba vytvo�it na microSD kart� v adres��i garmin soubor gmapsupp.img. K tomu by teoreticky m�l slou�it program sendMap. Bohu�el se mi nepoda�ilo doc�lit �sp�chu p�i pr�ci s mapami od Garminu.

Ostatn�

K update systemoveho software v navigaci je pot�eba instalace Windows �i Mac OS X.

�esk� datov� zdroje pro Tellico

Wed, 05 Mar 2008 20:58:00 CET

Aktualizov�no.

Tellico je vynikaj�c� spr�vce sb�rek pro KDE. Jedin�, co mi v n�m chyb�, je n�jak� �esk� datov� zdroj pro knihy. I napsal jsem p�r skript�, kter� umo��uj� vyhled�v�n� podle ISBN a um� i st�hnout n�hled p�ebalu:

Skripty jsou naps�ny v Perlu a ke sv�mu b�hu pot�ebuj� n�sleduj�c� moduly (v�echny se daj� st�hnout z CPAN):

LWP::UserAgent
IO::String
XML::Writer
HTML::TreeBuilder
Text::Iconv
MIME::Base64
Digest::MD5
Encode

Do Tellico se skripty instaluj� p�es Nastaven� / Nastavit Tellico… / Datov� zdroje / Nov�… s n�sleduj�c�mi parametry:

Druh zdroje: Extern� program
Druh sb�rky: Sb�rka knih
Druh v�sledku: Tellico
Cesta k programu: cesta ke skriptu
Parametry: za�krtnout ISBN

V n�kter�ch verz�ch je je�t� t�eba do pol��ka vedle ISBN zapsat ru�n� "%1" (bez uvozovek), a�koliv je tato hodnota ji� p�edvypln�na (pod�kov�n� pat�� Jakubu Fialovi).

Skripty mus� b�t spustiteln� (chmod +x).

O�et�en� chybov�ch stav� nen� valn�, ale pro b�n� pou��v�n� to sta��.

Java Mystery Of (Un)Closed Sockets

Mon, 03 Mar 2008 20:40:00 CET

Windows has a limit on number of existing client network sockets. I recently ran into this when I started to close sockets which were previously left to be timed out. That was certainly a suprprise, because naturally one would thought that closed socket will be, well, closed much earlier than unclosed one.

I created a small test case:

    public static void main(String[] args) throws Exception {
        int i = 0;
        while (true) {
            Socket sock = openSocket();
            OutputStream os = sock.getOutputStream();
            os.write("GET / HTTP/1.1\n\n".getBytes());
            InputStream is = sock.getInputStream();
            while (is.read() != -1) {
                // void
            }
            is.close();
            os.close();
            sock.close();
            System.out.println("" + i++);
        }
    }


    private static Socket openSocket() throws Exception {
        InetAddress address = InetAddress.getByName("localhost");
        Socket socket = new Socket(address, 80);
        socket.setReuseAddress(true);
        socket.setTcpNoDelay(true);
        socket.setKeepAlive(true);
        socket.setSoLinger(true, 0);
        return socket;
    }

The socket.setSoLinger(true, 0) should force the TCP connection to be dropped immediately with connection reset when socket is closed.

I tried that (and modifications without setting of SO_LINGER(0) and without closing the streams and socket) on Windows XP Professional connecting to Apache 2.0.59 and got this (the numbers differ between executions, but there is a general trend visible in a number of successfully allocated sockets):

	closed with `SO_LINGER(0)`	closed normally	unclosed
Java 5	less	more	more
Java 6	more	less	less

Fortunately Java 6 behaves more logically. But, of course, anyone allocating that many sockets should be punished.