Update Chinese sounds and background sound level

Add warning to README

Add Japanese sounds (language: "ja")

Contributed by email.

Fix typo in comment

Merge pull request #8 from rainyear/master

Update README.md

Use ServeContent in Server.

This makes Server support Range request, fixing issue with Safari not
playing audio.

Gofmt.

Fix modulo bias in randomBytesMod.

See
https://github.com/dchest/uniuri/commit/2e215562f605dacd124f15eb89523bdc32da3a41

capexample: listen on localhost.

In Server set headers to disable any caching.

Make sure dotSize is never less than 1.

This fixes panic due in PRNG "invalid argument to Intn" due to
requesting random number with no range (0).

Fix condition; remove unreachable panic.

Gofmt; make capgensound produce gofmt'ed output.

Generate captcha representations deterministically.

WARNING: introduces API incompatibility!

This package generates captcha representations on-the-fly; for instance,
if captcha solution was "123456", every call to NewImage() using this
sequence of digits would generate a different random image containing
"123456"; similarly, NewAudio() would generate a different audio
pronouncing the same sequence: 1, 2, 3, 4, 5, 6.

If a user, instead of storing generated outputs, exposes this
functionality from their server, which is the default and recommended
behaviour, an attacker could try loading the same image or audio over
and over again in attempt to arrive at the most correct optical/voice
recognition result.

Instead of using a global non-deterministic pseudorandom number
generator to distort images and audio, this commit introduces a
deterministic PRNG for each image/audio. This PRNG uses a combination of
a global secret key (generated once during initialization from a system
CSPRNG) and captcha id and solution to produce pseudorandom numbers for
each representation deterministically. Thus, calling NewImage() with the
same captcha id and solution at different times will result in the same
image (ditto for NewAudio).

To make results unique not only for different solutions, but also for
ids, these incompatible changes to public API have been introduced:

NewImage and NewAudio changed from:

  func NewImage(digits []byte, width, height int) *Image
  func NewAudio(digits []byte, lang string) *Audio

to:

  func NewImage(id string, digits []byte, width, height int) *Image
  func NewAudio(id string, digits []byte, lang string) *Audio

That is, they now accept an additional captcha `id` argument.
No other interfaces changed.

Described changes also improved performance of generating captchas.

Use SipHash-based PRNG for image/audio variations.

Merge pull request #2 from ebfe/typo

Fix typo.

Fix typo in comment.

Fix image.WriteTo, properly implementing interface.

Instead of writing directly via png.Encode, encode image
into a buffer first, then write the whole buffer.

~~~
This commit is brought to you by:

http://blog.oleganza.com
Must-have source of knowledge about Bitcoin.
~~~

(Sponsor my commits! https://github.com/dchest/commit-ads)

store: simplify locking.

Update .gitignore.

Update docs.

Add Chinese to generated sounds and example server.

Add contributed Chinese sounds.

Update README for capgensounds.

Rename generate to capgensounds.

Remove Makefiles. Rename example.

Go weekly 2012-02-07.

Example is now capexample.
To build helper tools:

   cd capexample && go install
   cd capgen && go install

This will install tools capexample and cangen into $GOPATH/bin.

Fix documentation.

Fix expiration constant.

Fix for Go weekly.2012-01-15.

Fix build for Go weekly.2011-10-06.

example: Use new template package; write text/html header.

generate: README fixes.

capgen: fix build, add "lang" flag.

generate: add info about forking to README.

generate: README fixes.

generate: add README about how to create sounds for new languages.

Remove notice about Chrome bug, as v12 has been released.

random: avoid modulo bias.

Add copyright headers.

sounds.go doesn't have a copyright header, because the file is generated
by a program. I don't want to make it output the copyright header,
because users can generate sounds themselves, and the copyright status
of such generated files can be different.

Better Russian sounds.

generate: use filepath instead of path.

generate: add new line after digitSounds.

Update README.

Mention Russian in package doc.

example: add language dropdown.

Add Russian sounds (ru).

Support other languages in audio captchas.

Server accepts parameter lang, e.g. xxx.wav?lang=en.
NewAudio and WriteAudio accept the target language.

Better error descriptions.

image: s/img/m/

Update example.wav.

audio: increase level of background sounds.

Update generate with new waveHeader.

audio: proper RIFF header generation.

At last!

[RIFF][following size][WAVE]fmt[...]...data[data size][padded data].

Also, update example.wav.

audio: error handling; pad byte is 0, not 128.

audio: write pad byte when data length is odd.

This is specified in RIFF format, but I forgot to implement.

     ckData  Binary data of fixed or variable size. The start of
	     ckData is word-aligned with respect to the start of the
	     RIFF file.  If the chunk size is an odd number of bytes, a
	     pad byte with value zero is written after ckData.  Word
	     aligning improves access speed (for chunks resident in
	     memory) and maintains compatibility with EA IFF.  The
	     ckSize value does not include the pad byte.

(From http://www.kk.iij4u.or.jp/~kondo/wave/mpidata.txt)

image: rename a few variables.

random: seed with Nanoseconds.

Remove dependency on uniuri. Add random.go.

Captcha ids are now 20 characters long.

Add store set/collect benchmark.

image: forgot to use circleCount const in NewImage.

image: use paletted image.

This simplifies code and makes it faster. It also produces smaller PNGs.

Before:

captcha.BenchmarkNewImage	500	 6198230 ns/op
captcha.BenchmarkImageWriteTo	100	17517110 ns/op   0.14 MB/s

After:

captcha.BenchmarkNewImage	500	 3869968 ns/op
captcha.BenchmarkImageWriteTo	200	 7237604 ns/op   0.23 MB/s

(Core 2 Duo @ 2.26 GHz, MacBook)

Currently the background is white, not transparent, because PNG encoder
doesn't support paletted images with alpha channel.

Submitted CL: http://codereview.appspot.com/4432078

Change alpha to 0x00 once it's accepted to make captcha transparent.

capgen: update for StdLength -> DefaultLen rename.

Report MB/s in audio and image benchmarks.

Also, audio benchmark uses the new ioutil.Discard.

audio: do not change final background level.

Instead, change level of individual reversed sounds when making
background to the appropriate values.

This speeds up makeBackgroundSound ~3%.

Before and after.
captcha.BenchmarkNewAudio      50  30436940 ns/op
captcha.BenchmarkNewAudio      50  29435780 ns/op

Also, level up white noise a bit.

audio: preallocate reverse sounds (~5% speed up).

Also:
- use ranges everywhere
- optimize makeWhiteNoise a bit

Before and after:
captcha.BenchmarkNewAudio      50  32128580 ns/op
captcha.BenchmarkNewAudio      50  30436940 ns/op

Add image and audio benchmarks.

Update README again.

Update README.

Fix tests.

image: remove random variations in drawDigit's x,y.

For the standard size and width with new font they no longer had any
effect anyway (were always +0).

image: precalculate dx in distort and strikeThrough.

server: better names for arguments to Server.

New now returns captcha with default length.

NewLen is now like the old New (accepts captcha length as the argument).
Also, rename StdLength to DefaultLen.

captcha: remove Collect method.

Garbage collection is an internal detail of a Store, and requiring to
export this method seem to be not very useful.

The default memory store, of course, still has garbage collection, but
it's not the unexported method, called in Set as usual.

image: fix terminology again.

image: refactor, fix terminology.

Update README again.

Update README.

Update example.png.

image: new wave distortion, and strike-through.

example: print the address of server.

font: more readable 1 and 7.

image: use font with better resolution.

image: simplify skew randomization.

Use rndf instead of rand.Float64 voodoo.

example: add "Try another one" link; ignore /xxx.

Only return page for root path /, return 404 for any other, such as
/favicon.ico.

server: fix "multiple response.WriteHeader calls".

The server no does nothing for errors other than ErrNotFound (for which
it calls http.NotFound). This is becase when getting error on writing,
we already wrote headers, so can't respond with status 500.

Possibly log errors in future?

example: update with recent download changes.

server: serve downloads from "/download/".

Also, refactor methods.

Move server parts to server.go.

Server: disable uppercase extensions (.PNG, .WAV).

Also fixes comment typo.

example: use io instead of fmt; add error handling.

store: fix potential race.

store: rename ids to digitsById, exp to idByTime.

Update docs.

Add examples captchas.

Use new GitHub-flavored Markdown syntax for types.

Rename README to README.md

Add README.

Example: make HTML5 valid.

Example: implement reloading; fix audio in Safari.

Server: reload captcha when given ?reload=x in URL.

Add notice about Chrome 10 bug to BUGS.

Add notice about Image.WriteTo to BUGS.

Remove "Std" from StdCollectNum and StdExpiration.

Add Store interface.

Standard storage is now known as memoryStore (created with
NewMemoryStore exported function). It's still default, however there's
now an option to replace it with a custom store by implementing Store
interface and calling SetCustomStore.

Add example HTTP server that uses the package.

Thinner strike-through on captcha image.

Add Audio.EncodedLen. Workaround for Chrome WAV bug.

Add VerifyString function.

Add CaptchaServer.

Fix comments, apply gofmt.

generate: rename to main.go, add Makefile.

Fix adding extra space in the generated sounds.go.

Rename cmd to capgen, originals to generate.

Add captcha test.

Add store test.

Fix collection (locking and getting next value).

newStore: accept collectNum and expiration time.

Prevoiusly, Expiration and CollectNum were hard-coded into store.go.
This change moves these constants to captcha.go (renaming to Std*), and
makes newStore accept them as arguments.

This will enable setting these paraments by package users sometime in
the future, if needed, and also allows testing of the store garbage
collection.

Add ErrNotFound, return it from WriteAudio/Image.

Export RandomDigits. Remove NewRandom* functions.

Rename "numbers" to "digits" everywhere.

Plus fix comments.

Set capacity for audio buffer. Rename makeStaticNoise to makeWhiteNoise.

Rearrange code and fix docs.

Implement a single captcha test command.

The new command replaces old test command-line utilities from cmd/image
and cmd/audio directories. It can now generate both image and audio
captchas, and accepts a few useful flags.

Edit .gitignore to reflect changes.

Add NewRandomAudio function.

Audio test cmd now writes to stdout.

Replace PNGEncode with WriteTo for Image.

Add WriteAudio function.

Rename CaptchaImage -> Image, CaptchaAudio -> Audio.

Apply gofmt; fix comment in generate.go.

Better randomize background and sounds.

Shorten intervals between prelude beeps.

Lower level of beep sound.

First version of audio captcha.

Move image generation test program to cmd/image.

image: move border calculation to calculateSizes().

Fix comments.

Move rand initialization to image.go.

Clarify docs.

image: refactor; add more color variations.

Refactor store into its own file.

Accept the length of the random sequence of numbers.

Call Collect on a separate goroutine.

Update documentation.

cmd: use standard width and height for image.

Document functions and types in captcha.go.

Refactored image captcha ready, and documented.

Numbers are now laid out correctly.

Next commits will introduce perturbations.

Started work on code cleanup.

Gofmt font.go.

Move image-related functions into image.go.

Add line that strikes through digits.

Reorganize code. Add storage.

Initial version (numbers).

Fix typo.

Update README.md