Antarctica is quite a bit different from other continents due to the climate and the resulting abundance of ice. This page describes some of these differences and possible ways to represent them in OSM data.
Most of Antarctica is covered by ice up to the coast, therefore ice free areas are the exception rather than the rule. The ice free areas none the less need to be mapped. Because it is practically unfeasible to create a huge multipolygon for the antarctic ice sheet with tens of thousands of inner rings representing the ice free area it has been decided in 2013 that land areas south of 60 degree south - including islands (this line being the 'political border' of Antarctica and also a reasonable logical division) - are assumed to be covered by ice, i.e. are meant to be treated like natural=glacier.
Practically that means the land ice does not need to be mapped explicitly - though mappers can add natural=glacier for dedicated mapping of individual glaciers. Glaciers or ice streams within the ice sheet without verifiable boundaries can also be mapped with linear ways or nodes.
Ice free areas need to be explicitly mapped to be considered as such - practically that is mostly natural=bare_rock and natural=scree. This in particular also applies to ice free islands which - in addition to the coastline - need either a second way or a multipolygon relation with the coastline as member tagged natural=bare_rock. Otherwise they will be and should be interpreted to be ice covered.
Rather than having a huge multipolygon for the ice it might be a good idea to consider ice to be the default land cover for Antarctica and map the ice free areas individually as polygons (mostly natural=bare_rock, natural=scree). This would require support from the renderers. Individual glacial features could still be mapped (for example to name glaciers) by drawing them like elsewhere above the default ice.
The ice shelves around Antarctica are thick plates of ice (thickness of 100m and more) which exist at 44 percent of the antarctic coast. More details can be found on Wikipedia. How exactly this looks like in principle and where the various lines talked about in the following are can be seen in the image on the right.
Consensus in OpenStreetMap is that the coastline is to be mapped at the outer edge of the ice shelves (the calving line) even though the ice shelves are floating on sea water. We do that because this is the logical coastline with regards to human activities. Typical ocean use (i.e. ships) can only take place outside this line and typical land use (like buildings, airfields etc.) is fairly unrestricted on the ice shelf. Ice shelves are not simply thin floating ice, they are thick and stable over decades and more. Also see Sea ice below.
The ice shelves are represented as mutipolygons and are - corresponding to Proposed_features/Glaciers_tags - tagged glacier:type=shelf. Map rendering should best show them in a distinct way, different from normal glaciers.
In addition the outer edge (the calving line) and the inner edge (the grounding line) get different tags for the corresponding ways:
Sea ice is relatively thin, much thinner as the ice shelf and it is seasonal, so it changes with the time of year. Around the North Pole you have only sea ice, but not ice shelf. Sea ice is not mapped at all in OSM because it changes all the time.
Antarctica islands are called islands no matter if they are in the free ocean or enclosed in ice shelf. Their perimeter (ie the way) should be tagged as place=island in both cases (in case of islands enclosed in the ice shelf the edge would be the grounding line). If and only if they are in open water they are also tagged with natural=coastline. Like with all islands a node with place=island and a name can be put in the middle of the island.
Long straight lines are problematic in OSM everywhere. The earth is round and depending on the projection used for a map those straight lines should sometimes be rendered as curves. But the computer doesn't know that and will still just render a straight line between the two end points. This problem is not specific to Antarctica, but it is seen there more often than in other places, because map distortions are often greater and because features are often drawn with less detail than in other places in the world.
To mitigate this problem you should split up long straight lines every 100km or so with extra nodes.
There are certainly further features which are rare or non-existent on other continents. The above are only the large scale phenomena that will be important for the import. Feel free to add other things here to help develop uniform standards for mapping.