Yesterday I attended the WebXpo event in Sofia, managed by Web & Events. There were plenty of different topics to be discussed on website development, design, usability and so on. I've written a review in Bulgarian in my other blog.

It was a great pleasure listening to the discussion of the website price. Three experts from leading web development/design companies in Bulgaria did a workshop on how prices are calculated, how do they manage different clients and complex tasks etc. All of them mentioned that the higher prices when you recall to company are based on the professionalism that they imply in their products. The whole cycle requires brief specifications, development and design, testing, some usability experience tests, support, SEO and others. Things that could not be accomplished entirely by a small team of non-qualified and experienced developers. This one is a bit tricky, of course, because the companies have higher rates as well and disciplined way of working when the client might need some flexibility at this point. Anyway.

Some of the hot quotes from the event:

"If you have a website for yachts, you don't need 10K visitors, but 1 buyer."

"When it comes to support and fixes, relatives suddenly disappear"

"A website could be developed by a company or 2 students in the Technical University, but the probability it would ever be deployed is way higher when working with the company"

"How do you manage changes to the specs?", "We call the client with the new proposal (changes to contract and price)"

"Content has to be provided by the client, not the company. Otherwise he has to hire copywriters as well."

"10% of the people use IE6. This would increase the final price with 600lv. You want this or not?"

I've been researching the Django framework for the last 2-3 weeks in my spare time. It's development is going on and on and I see many interesting innovative concepts realised behind.

I tried it 3 years ago for a course project when we had to integrate the PIL (Python Image Library) into a web project with the existing technology. It was still first version and not much functional though. Poorly documented as well, but as I see now it is almost complete. Python is a powerful language and a web framework based on it i s a great idea.

There are few CMS systems based on Django. Honestly I don't see any sense in any of them - basically because they could save you 4 or 5 hours of work total but with something that could also trick you later. The admin generation is fast enough in Django that you don't need a few days to moderate the administrative panel and do all the necessary relations for the CRUD .

However, there is one thing I do not like in the basic concept. Comparing Django to CakePHP (which has similar abilities in it's latest releases) I feel restricted by the automagic thing. Automagic is a code-behind technique that generates no code in your project but does stuff at the same time. The problem with this all is that you have no idea what happens behind the project itself. In CakePHP you have admin generation in the same way you see it - yes, 'generation' with 'code generation' as well - files + tests are done and added to the project, controllers have their own methods that execute functions and so on. When you change function code, the project acts a different way. In Cake there is automagic as well, for instance if you define function login() {} for the auth module, you already have login system. No fields defined, nor actions. Just an empty login function. But this is more of an exception than to a rule.

When you define a Django project, the admin panel has no fully coded Admin classes. Everything is reused from the parent classes and once you need to change, you need to research either the documentation or the code in order to find the required methods. This is hard to find in the beginning of your Django career having no expectation of the backend and the possibilities of functions for the project.

Anyway, if we agree on a bit harder learning curve, the framework itself give much more than coding from scratch or with any other framework. Powerful applications could be defined if not in hours, in days, and could be deployed in places like the many Django hosting companies abroad. In Bulgaria there are host providers that support Python (via CGI) and install Django framework on request.

Three things with not much in common, except that these are the three interesting things around me these days.

For the last 2 weeks I do python development (actually Django) for an international project while I'm ending my CakePHP and Digital Signer applications. I set up Subclipse in my Eclipse because I got tired of committing twice some of my files. This way my SVN is connected to the project in the IDE and the new data is deployed in an easy and beautiful way without any console commits.

The Django projects are fine as well. I do researches over the multilingual applications. So far I found url-locale and multilingual modules which serve to detect the language via the URL and create a multilingual content while create the intermediate tables for that. I continue researching the session system in order to save the languages there where switching around.

As for the Android, I plan to do my first app for almost a year. Recently I had a few job proposals for Android development plus the Mtel contest for Android app. That's why I downloaded the SDK and the Android plugin and started from the Hello World tutorial online. More researches are done and I am willing to test a more complex application with Internet activity, of course.

We delivered Motoclub.BG and now it's available and accessible. The web portal is an online version of a magazine for motorcycle sports and vehicles, ATVs and others. As the bikers' favourite music is rock/metal, breaking news on concerts and live events in clubs are posted there as well.

The portal has few interesting sections for magazine online overview. You could find news and specialised articles, moto calendar with coming events, catalogue with specifications for different brands and so on. Filters and archives provide searching via keywords or month/year ranges. RSS feed on latest news is staying in the left column, next to the invitation to the Facebook group of the website.

The project is CakePHP based with MySQL database behind. I've implemented a few custom solutions on user management and RSS aggregation which I'm going to describe later.

I had a training on freelance tips and tricks in New Bulgarian University on Thursday. I uncovered some of the issues I've had during my first years of freelancing as well as some tricks how to avoid suspicious clients and projects.

My presentation included:

• what does freelance mean
• how to switch to freelancing
• where can we find projects and clients
• how to interact with them
• how to present ourselves the right way on/offline
• how to define our costs
• how to manage the time
• etc

Slides in Bulgarian are available in my Slideshare account.

It's an algorithm that could be applied almost everywhere and requires no special skills. First of all, lets define the cons and pros for you so to be able to chose whether you would get the advantage or not.

Cons

• depends on the particular site at some point (changes dependant)
• requires few seconds for thinking (depends on complexity of your own algorithm)

Pros

• absolutely and undoubtedly unique algorithm. The base is clear, the implementation is on your own.
• could be extended or tuned if needed - flexibility is everything
• no need of keystore or Internet connectivity
• hardly guessable if someone steals one or two of your passes

If you count on the pros, there we go.

The trick

 This password algorithm is similar to the ones in the websites. Where you have:

a) domain name

b) keyword of yours

so the website hashes the domain with your keyword and generates a unique string literal to be used as a password value. The problem stands in the unreliability of the websites - you might not have Internet connection, or it could be change, or someone could break through or something.

Our algorithm works that way:

1. you take some areas of the website/program (domain name, header, title - something not willing to change soon)
2. you crypt it on your own
3. you use the final sequence as a password
4. as a result, every final sequence for each site might look different and hardly guessable

The main issue here is that you count on the site/software to be constant at all. If you rely on the domain name and the owner migrates to another one, you have to use the 'forgotten password' link or just remember the last version change. Shitty, but rare case.

Examples

Few examples for prove of concept.

Simple Algorithm: we count the domain symbols (length), multiply by 3 and append the number to the reversed website name.

If we take for instance google.com, then our password would be elgoog27 (google reversed and the length 9 multiplied by 3).

When using freelancer.peshev.net, there we have recnaleerf57 (if we use only the first part of the domain as algorithm).

Complex Algorithm: we get the second and third letter of the website, increment the first letter alphabetically ('a' becomes 'b', 'd' becomes 'e' etc.) and decrement the second. We use the string '$xZ' as a constant after the transformed letters. We append the last 3 symbols of the source code of the main page. At the end we add the length of the subdomain (if any), the main name and the top-level domain.

Taking the same example google.com, here we have: pn$xZpt>63 ('oo' with inc(1) and dec(1), then the literal, then the last three symbols of the source, which is script tag and finally 63 (length of 'google' and length of 'com'). Hardly guessable, although that the source code might change at some point (use at your own risk). 

If I apply it to my own freelancer blog - freelancer.peshev.net, I would have sd$xZml>1063 ('re' translated, the constant, the html tag end and the length of 'freelancer', 'peshev' and 'net').

...

You could define an algorithm of your own, based on your own needs and expectations. The point is that you remember only the steps, the rest is some translation on runtime, when you enter the site.

If you need to create a complex regular expression, the best you might do is to debug it realtime onchange. One of the best free online tools for that is the REGex TESTER.

Via Regex Tester you might type a sample text and regular expression and dynamically see the matched substrings in few regex pattern types. Some flags are modifiable as well. At the bottom of the page you could also find some of the most popular regex conditions - email pattern, IP pattern etc. to be used for validation purposes.

I am working on a project that requires PDF signing through a website.

The current infrastructure includes Apache + PHP website available. What we need to achieve is integrate the digital signature module of a user visiting the website and sign the required files. One of the most appropriate solutions after the negotiations was signed applet that signs PDFs.

1. The steps for the project creation are the following:
2. Create an applet
3. Create a certificate to sign the applet via keystore tool (if root CA's are not available)
4. Sign the applet (allowing read/write access to users' hard drives)
5. Create a REST implementation that could read/write to the server database via HTTP
6. Implement the iText API
7. Create an appropriate interface to list certificates and documents, too
8. Sign each of the documents

Some useful materials on the topic:

How to sign a Java applet

How to sign an applet (and get it to work)

Oracle Signed applets tutorial

Creating PDF with Java and iText

How to sign a PDF using iText

LifeCycle ES PDF signing (the Adobe way)

If you use Oracle for your projects, you are probably using your custom made interface for database connectivity, SQL*PLUS (integrated in the Oracle package) or some external 3rd party tool.

My choice on DB access when Oracle is the choice is Oracle SQL Developer.

Some features of this tool (from Oracle website):

1. Create Connections
2. Browse Objects
3. Create Objects
4. Modify Objects
5. Query and Update Data
6. Export Data and DDL and Import Data
7. Schema Copy and Compare
8. Process Commands
9. Edit PL/SQL
10. Run and Debug PL/SQL
11. Run and Create Reports
12. PL/SQL Unit Testing
13. Data Modeler Viewer
14. Migrate from Third Party Databases
15. Version Files
16. Review the Robust IDE

One of my favourite functions is trigger on sequence creation. When we need auto increment facility for a table, we usually pass through the following steps:

1. Write the table code
2. Write the sequence code
3. Write the trigger code, connecting sequence to a 'before insert' action on a table.

Using Oracle SQL Developer, we could do:

Create a table with few clicks (defining the appropriate fields)

Create a sequence with the parameters in predefined form from the tool.

Then right-click on the table and create a Trigger -> PK from sequence which would create a trigger, catching a sequence and auto-incrementing an ID key of our own choice.

Great job that decreases the code writing when not necessary.

Last week I gave a database training in Switzerland. We had a look into relational databases, normalization theory, some good principles on database modelling, SQL language.

As of the used tools, we had an overview of Oracle 11g free database server, as well as Oracle SQL Developer tool to access the database systems. The E/R diagram toolkit we used to train on was TOAD data modeller. The previous versions of the tool were freeware, but the last implementations available are only evaluation for 2 weeks. You could try to find an old version that does the general work - database drawing, generation of SQL (as well as simple reverse engineering), creation of relations - identifying and non-identifying, as for one-to-one, one-to-many and many-to-many relationships.

In my training here I also extend the presentations with PL/SQL tips and tricks, Object relational mapping with Hibernate and JPA and few practical projects in classes.