Intro
The Data Basket UI design shows shortened URLs (eg "RS.ly/MK3CD").
Such functionality was not discussed in the spec, but was accepted as a given during the UI design discussion.
Here we will put the design decisions about how we'll implement URL shortening.
Disadvantages
Shortened URLs have many disadvantages including unreadability, reduction of trust, instability, overhead traffic, and security risks.
- http://rield.com/faq/why-url-shorteners-are-bad : an excellent page discussing all aspects of shortened URLs.
- http://sebastians-pamphlets.com/category/s-url/ : an emotional article including numerous links to other pages criticizing this practice.
Includes phrases such as- Twitter decided to break the Web by raping all of its URIs. Twitter’s sloppy URI gangbang became the Web’s biggest and most disgusting clusterfuck in no time.
- What if the great chief of Libya all of a sudden decides that gazillions of bit.ly-URIs redirecting punters to their desired smut aren’t exactly compatible to the Qur’an? All your bit.ly URIs will be defunct over night.
- Goo.gl (by Google)... promises stability, security, and speed. Well, at the day it launched, I broke it with recursive chains of redirects, and meanwhile creative folks ... wrote a guide on “hacking goo.gl for fun and profit”
- http://sebastians-pamphlets.com/put-an-end-to-uri-shortening/
- http://www.kottke.org/09/04/url-shorteners-suck
- http://joshua.schachter.org/2009/04/on-url-shorteners.html
- http://tag.us.com/uri-shorteners-suck-ass.htm
- http://techcrunch.com/2009/04/06/are-url-shorteners-a-necessary-evil-or-just-evil/
- http://gregable.com/2009/05/why-do-we-even-need-url-shorteners.html
- http://searchengineland.com/time-to-think-carefully-about-which-country-hosts-your-url-shortener-52579
The fact that you can surmise what the above pages are about is a simple example why shortened URLs should not be used
Software
References
http://rield.com/faq/why-url-shorteners-are-bad :
- the penultimate section describes Free & Open-Source URL Shortening Software.
- the last section has Further Readings, including
Free URL Shortener Comparison Spread Sheet (spreadsheets.google.com)
URL Shorteners: Which Shortening Service Should You Use?
While doing evil, reluctantly: Size, er trust matters
How to cleverly integrate your own URI shortener - speed and reliability comparisonhttp://royal.pingdom.com/2010/10/29/is-goo-gl-really-the-fastest-url-shortener-chart/http://gregable.com/2010/09/googl-url-shortenere.html
Stateful or Stateless
The key question is whether to:
- use a state-full service (i.e. save the redirects),
- or a functional service (i.e. use a 1:1 mapping based on a hash, with no need to store state).
Since we won't let the user select the short URL (a "vanity keyword"), it's better to use a 1:1 mapping
Internal or External
Should we deploy our own service, or use an external service?
I think internal, because:
- see the explanations about security risks and instability
- we'd need to access the external service somehow, and perhaps maintain the mappings
- such service is not very complicated
If external, which one should we select? I'd go with goo.gl
Domain, Resolving
If we decide to go with internal software, then we need to get an actual domain, establish DNS record, and setup a redirector. This is required for short URLs to be resolvable (to work).
Examples of shortening domains include:
- youtu.be
- tcm.ch
- twurl.cc
- tinyurl.com, snurl.com
- is.gd
- goo.gl
- cli.gs
- tr.im
- binged.it
- bit.ly, ow.ly
- fb.me
- su.pr
- twl.tl
These are Top Level Domains (TLD), belonging to some country. Eg bit.ly is Lybia (silly!) and goo.gl is Greenland.
To get such domain, one should follow the specific procedures of the respective TLD authority.
Alternatives:
- The UI mockup uses domain "RS.ly" which is modeled after "bit.ly" but the abbreviation doesn't make much sense.
Also, I don't think it's a good idea to deal with Lybia - We could get a domain such as prj.rs (stands for "researchspace project"); rs is the Republic of Serbia TLD.
Could also use subdomains per project, eg rembrndt.prj.rs, bm.prj.rs, etc. - If we select goo.gl:
- is it ok if our URLs are random, i.e goo.gl/XXXX
- maybe we could get and "reserve" a prefix, so our short URLs are goo.gl/rsXXXXXX or goo.gl/rs/XXXXXX
TODO Dominic: Decide about domain
Design
RS-954
TODO Ceco:
- research and trial some open source software (java!)
- research goo.gl and see how RS could use it (what is the API?).
Is it possible to "reserve" a prefix? - describe how we'd deploy the sofware, or use goo.gl
Possible solution that I found and tested:
- tinyURL Service
- easy for use and deploy
- open-source
Deployment
1.1. Download the javaQuery API
http://sourceforge.net/projects/javaqueryapi/files/javaQuery1.0/javaQuery%20Bundle/
1.2. Import downloaded jar to the project build path
1.3. Invoke
tinyURL tU = new tinyURL();
tU.getTinyURL("http://blog.wireframe-studio.com/wp-content/gallery/avatar/avatar-neytiri-movie.jpg");
1.4. Result
2. Goo.gl Service
Deployment
2.1. I found a simple class that changes the given String Object (the url) to a short and easy to remember Goo.gl URL.
To work we need to download GSON library - http://code.google.com/p/google-gson/downloads/list
(gson - convert Java Objects into their JSON representation)
2.2. Add the main jar to the build path – the test was made on 08.14.2012, main jar was
gson-2.2.2.jar
2.3. Import in the project next java file that I upload (Google.java)
2.4. Invoke
Google.shorten("http://www.premiumpress.com/forums/topic/website-thumbnail-api-premiumpress-free-api-or-disabled");
2.5. Result
Note: From result 1.4 and 2.4, shortening url(SU) working but the prefix is based on SU service provider, we can’t reserve ours;
The best solution is to create own url shortening service, that way we can reserve our domain.
3. Creating custom algorithm for shortening urls
The idea:
Suppose we have a table with following columns:
- unique auto increment id (long),
- url (string),
- base62 string (string)
The trick, convert unique id to base62 string not the url, and then the url is mapped to the unique id
Base 62 converter code uploaded !
Storing and Processing
TODO Vlado:
- Need two properties for full and shortened URL.
- display the full URL in tooltip
- describe that RS URIs are not resolvable yet, so maybe we should use a fixed short form such as "uri.rs"