Changes from version 0.3.6 to 0.4-beta1:

Major changes
Basic image support added
gzip content-encoding supported
Cache files to flash card (mostly optional - see config.ini)
Minor changes
issue 32 - file that defines allowed cookies can now contain http:// at the start
Plain text rendering semi works again (no font changes though)
Unicode shrunk to 16 bits internally to use less memory
!URIs with spaces are escaped to use "%20"
libnds-20071023 compatibility changes
Build changes
Build with devkitArm r21
Changed to scons build system

Bug fixes in version 0.3.6:

Fix newline bug in header parser for servers that send \n instead of \r\n
Remove the accidental inclusion of cache dir ( issue 33 )

Open the champagne !!!
I've here an input plugin which is named DsPad and uses SDL.

Enjoy, because it seems Lino's directinput plugin is buggy, it displays a message saying "Error -8" and all of its buttons in its control panel are greyed.

DsPad supports up to 4 joysticks and offers a complete joystick support.
No worry, the latest actual version of the SDL (1.2.12) is provided.
I hope new version 1.1.0 will fix bugs, I changed some things, I think the conflict is solved.
New version 1.2.0.
New version 1.3.0.
version 1.4.0 has been released by Squall-Leonhart, he just changed 2 things : reorganized GUI and put OpenSDL on DllMain, it causes iDeaS to hang up on launch. He didn't solve the bugs. I experienced myself the bug with Y button.
New version 1.5.0, Will definitively solve bugs, inspired from source of Lino's DInput Plugin. Reorganized GUI, added support for POV diagonals.

First place SDL.dll in the iDeaS root folder (not the plugin folder)and DsPad.dll in the plugin folder.
Launch iDeaS and select DsPad in the Input plugin list. A file named DsPad.ini will
be automaticly created in the iDeaS root folder. It is the config file of DsPad.
Now, open the Property page and you will notice a tab named "DsPad Plugin Config".
This is the config panel of the plugin, here is its description :
All the buttons in the "Buttons config" frame are for button mapping. Click one and
press a key, a joystick button, a joystick hat in a direction or move an axis and
the selected button will be bound to the control you moved.
In the "Misc" frame, you can choose the axis sensitivity.
-Low : for very sensible axes.
-Medium : recommended.
-High : for very used or lowly sensible axes.
The config is saved each time you modify a value, but you can click the "Save config"
button to make sure the config is saved.

Today's experiment... FAILED as well.

A very nice person calling him/herself Yuki donated Japanese ROM set, fresh from Dreamcast. Thank you! To my suprise it's 1.01d, just as any EU/USA one... And the FLASH file is a bit different too, comparing to what I had until now. Some version strings from both BIOS images:

KABUTO Ver.1.004
GDC Version 1.01 1998-09-30 MP
syTmr Version: 0.51 / syTmr Build:Jun 23 1998 14:52:57
syRtc Ver 1.00 Build:Sep 28 1998 05:10:44
syInt Ver 1.00 Build:Sep 16 1998 14:37:57
fmMng Ver 1.00 Build:Sep 29 1998 14:42:44
SIMPLE PLAYER VER. 1.00 98/10/02

KABUTO Ver.1.01d
GDC Version 1.10 1999-03-31 MP
syTmr Version: 0.51 / syTmr Build:Jun 23 1998 14:52:57
syRtc Ver 1.00 Build:Sep 28 1998 05:10:44
syInt Ver 1.02 Build:Oct 08 1998 19:31:47
fmMng Ver 1.00 Build:Sep 29 1998 14:42:44
SIMPLE PLAYER VER. 1.00 98/10/02

Well I can't be 100% sure, as I don't really know where that 1.004 came from, but it seems there was a bug in GD handling routines. GDC module was clearly upgraded. See, the older one is a bit strange, it's trying to boot GD-ROM media even if the GD drive says it's empty. This causes BIOS to hang just after swirl logo and you can't enter console menu. There is a way around it - namely you need to use unbootable image or (preferred) GDI image for EU/USA region rather then leaving the drive "empty". This will satisfy GD library and wont boot at the same time, so BIOS menu becomes accessible. I have no idea if that's because the ROM image was modified by someone or if it's a bug in the emulator. It is also possible that the early Dreamcasts were just slightly bit different then the later ones. Maybe there was even some sort of hardware-based workaround for that BIOS or the GD firmware was different.
The point is: If you use Japanese BIOS on Makaron and are experiencing problems with 1.004 version, substitute it with EU/USA image. This might not work well with the earlier FLASH image, but it should in most cases. The earlier FLASH lacks second entry for default settings (which include TV output format), but as far as I can tell it's not really used in NTSC modes anyway.

Oh, and I forgot to make an update to my last entry, so here are some final images from the experimental depth-peeling rendering system:

In some cases it simply looks great and most of the Z-fighting is gone, but I still couldn't get around all accuracy issuses. The last image is 7-layers deep and should render text in that dialog window, but...
Also, this is way too SLOW. For example (if you want things blended properly) to render 5 layers mean 10 redraws of the entire translucency. Fourteen redraws in Evolution is still real-time, but some games are way more demanding. For example my rig (E6600+GF7900) simply cannot handle more then four layers in Ikaruga and still maintain 60RPS, and in this game even eight layers is nowhere near enough.
I'm not going to include this experimental rendered in the next release, it's simply way too... experimental Even for Makaron.

It seems I suffered from a bit of memory loss while implementing the Amnesia effect, allowing it to be released in a rather half finished state. In 089 Soul Suckers will have their proper attack working. And also not look like floor tiles.
From external testing, reworked tutorial to make it a bit more friendly.
Dig no longer has force bolt as prereq, mana cost reduced to 10.
Spell book of utility broken into book of force and book of earth.
You see a scroll of READ ME now properly has the "a".
Character dump has skill and spell list. (R. Dan Henry)
x, +, * sources used in chararcter dump to show source of intrinsics/spells.
Soul Suckers will no longer try to cast at people they can't sense.
Amnesia shows up in your intrinsic list and status line.
You can no longer cast spells when suffering from amnesia.
New mob, gold beetles.
When a creature made of X attacks without a weapon in melee a creature vulnerable to X, extra damage is properly applied.
Tamed creatures will not continue to attack newly tamed creatures. (hotpoo)
Items of poison resistance can no longer be poisoned by dipping in a poison potion, you now get Nothing Happens.
Creatures that random poly will not random poly into themselves. Specifically, this prevents chameleons from being polyed as chameleons. (Clare Boothby)
I think the holes to bypass Belweir's test have been plugged...
I knew I'd scream when I hit the 256th tile. I was thus surprised when it (the signpost) worked so well. Needless to say the Soul Sucker showing up as a floor tile came as a surprise to me - I still had a bug with greater than 256 tile indices for items and monsters!
Surprise now has two rs in three affected messages. (David Damerell)
Fix a display issue where plain robes would report 5 AC when no skills were known. This should have been 3 AC, thereby avoiding the apparent reduction to 4 AC when a skill was learned. The actual effect was correct, just the display, and just the display of items with base ac of 1.
The plural of Guide to Armour is no longer Guide to Armours but the correct Guides to Armour.
Tlosh no longer gets upset at Reclaim Soul being cast. While it can be considered a type of healing (so discouraged), the method of healing is sufficiently grisly that Tlosh approves.
The breathe command provides better results when one is strangled/submerged/breathless.
One new room.
Leech the right version:

Gameboy Advance [375k].
Nintendo DS [725k].