I’ve written loads about postcodes before. GPS Ireland have published a concept for a postcode system, and Garmin have implemented it in a limited way on their devices. Good for them for publishing their idea.
However, there are a number of big issues with implementing such a system.

– The problem is that it is not much help in and of itself to actually deliver anything, except maybe by helicopter. You need to cross-reference the code against a map database to have any idea of how to effect the delivery or to estimate how long the delivery might take. This means that you are in a proprietary situation right away. There is no free road map database of Ireland.
– Packets and parcels with the code (and packets and parcels will represent 80 percent of delivery items within 20 years) would be practically impossible to sort by hand. You have to key ’em or scan ’em to have any idea which bag to put them in. Manual sorting of letters by unskilled workers is still a reality of the Irish postal system at the busiest times. Many other items (like registered letters and packages) are also manually sorted.

– You can have very ambiguous codes. For example, if you have a short back garden, you could end up with a code the same as the guy in the house backing on to you. In some cases, it could be a one-mile drive to the other property.

– The resolution of the system is just not that high for central areas. It’s only 5m. A significant number of houses in Ireland are smaller than 5m on one or other dimension.

– You really need a centralized database of addresses to do most useful tasks with the code (for example, verifying an address as valid). Sorting out this national address database is the big challenge for any postcode implementation. To do the surveys required from scratch for this system would cost at least EUR 15m, and maybe more.

– Google maps is just not an accurate way of determining longitude and lattitude. It is often many inaccurate by tens or even hundreds of metres, because of the way the projections work.

– letters in codes are not a good idea. They make it more likely that the code will be misremembered, misheard or misread. Letters that look alike, or like numbers, such as  ‘S’, ‘Z’ and letters that sound alike, such as ‘N’ and ‘M’ in particular are going to cause confusion and errors.

My own suggestion, which I have worked on with Michael Everson, is to use an all-numbers 5+4 code and base the code on existing administrative divisions, subdividing to road/street and individual delivery point level as necessary, and to number the codes geographically, i.e, the district in the top left of a county is numbered 101 and the district in the bottom right is number 909.

This has the following advantages:
– keeps the implementation costs to a minimum at the outset

– allows items to be be easily sorted manually or by machine

– can be implemented in a basic way without any proprietary databases (although it can conveniently link to proprietary databases)
– integrates well with existing databases (like census, rates office, electoral register and geodirectory)
– does not absolutely require a single address database at the outset (although one may have to be constructed over time on the basis of existing databases)

– allows the system to be extended where necessary (subdivision to provide a unique address for each home is not a big issue in Dublin City for example, but it is a big issue in some rural areas and is more likely to be cost-effective there).
– can cope with certain areas of the country becoming much higher density over time

– makes it obvious for a non-specialist withough special equipment as to where in the country a particular address is located.

This scheme isn’t perfect either of course, but after talking to a lot of people and considering a lot of options, it seems like the best way to do it. We also have a plan for giving the postcode legal backing, without requiring an Act of the Oireachtas.