Dung,
That excerpt you posted makes a good point about the vunerability of low-level readings to local disruption (especially during the war years in Geissen, I imagine).
I have not dug into the details behind the long article by Tony Brown. He claims that accurate measurements were possible from the early 1800s onwards and that many were taken by rigorous scientists who were aware of the possibilities of contamination. He mentions a set of high readings taken in Greenland which were attributed to volcanic activity, for example. Here is his description:
"Around this time Cavendish made some 500 samples of air by nitric oxide eudiometer and de Saussure took daily measurements for 3 years, this evolved into the more reliable hydrogen eudiometer. These were very accurate and those scientists taking measurements from around 1800 were well aware of the importance of geography, weather, wind, season, altitude, contamination etc when taking a reading.
CO2 readings from 1790 to 1820 should be considered interesting (and possibly approximately correct) but it is from 1820 onwards that the level of reliability increased enough for us to consider a meaningful proportion of them as a useful record of their time and place. In examining a few of the measurements taken at the time later in this article, it should be borne in mind that they are a fraction of many hundreds of thousands of independent readings taken by many scientists-several of them Nobel winners-from around 1830 to the advent of readings at Mauna Loa in 1957 by Charles Keeling."
From another angle, the studies of plant proxies show widely varying levels of CO2 in both the recent and more distant past. In particular they show some rapid swings on decadal and century time scales which are not reflected in the ice cores.
The WUWT post by David Middleton includes a possible reason for that discrepancy by explaining why ice cores give good resolution for temperatures but not for CO2.
For the moment, until something more convincing comes along, it looks to me that there are grounds for doubt over the "hockey-stick" profile of CO2 levels touted by climate science. I was hoping we might hear from Richard Betts on this - especially the reasons for discounting contrary evidence - but I expect he is somewhat distracted at the moment.

Dung,
Your comment puts things in a different light. My understanding of the ice-core measurement techniques has always been a bit vague. I was under the impression that the temperature info came from the ice itself, not the trapped gas, via analysis of the oxygen isotope ratios in the ice. Thus it was tied much more precisely to the date of the intial snow deposition.
Whereas the bubbles - I understood - are prone to vagaries since they form over a significant length of time, tend to migrate, have possible gas/water/ice interactions at the boundary, etc..
Rats! Now I need to learn more about the cores as well as the historic CO2 measurements.
However I do have a high regard for victorian scientists and their achievements. Like you I am open to learning more but natural variation looks more likely to me than steady-state.
It also strikes me as ironic that the IPCC fought so hard to defend the proxy-derived temperature "record" but, afaik, discarded the equivalent, proxy-derived info for CO2.

A post on the AR5 thread included this link:
http://wattsupwiththat.com/2010/12/26/co2-ice-cores-vs-plant-stomata/
It is another article by David Middleton. In it he pulls together the main sources for CO2 records and compares their attributes. It is well worth a read but here are the conclusions:

Ice core data provide a low-frequency estimate of atmospheric CO2 variations of the glacial/interglacial cycles of the Pleistocene. However, the ice cores seriously underestimate the variability of interglacial CO2 levels.
GEOCARB shows that ice cores underestimate the long-term average Pleistocene CO2 level by 36ppmv.
Modern satellite data show that atmospheric CO2 levels in Antarctica are 20 to 30ppmv less than lower latitudes.
Plant stomata data show that ice cores do not resolve past decadal and century scale CO2 variations that were of comparable amplitude and frequency to the rise since 1860.

Thus it is concluded that:

CO2 levels from the Early Holocene through pre-industrial times were highly variable and not stable as the Antarctic ice cores suggest.
The carbon and climate cycles are coupled in a consistent manner from the Early Holocene to the present day.
The carbon cycle lags behind the climate cycle and thus does not drive the climate cycle.
The lag time is consistent with the hypothesis of a temperature-driven carbon cycle.
The anthropogenic contribution to the carbon cycle since 1860 is minimal and inconsequential.
I have not had time to do any more digging - this just happened to catch my eye. It would be interesting to read why the non-icecore sources have been disregarded: can someone explain or provide links?